Every error this guide has fixed in a published fact, newest first: what the page
said and what it says now — and, on the entries that have one, how the error was found.
Reporting an error
To report an error, email
corrections@weedvoter.com with the page address and what you
believe is wrong. A source helps, but is not required.
Every claim on this site links to the document it came from, so a report can be checked
against the same source. Where a report is right, the page is corrected and an entry is added
below. Corrections are never made silently.
The log
98 corrections. Entries from before September 22, 2026
were reconstructed from the project's dated working notes, and smaller fixes from that period may
not appear. From September 22, 2026 on, every correction to a published fact is
listed here.
Four no-record cards listed pages as read on sites that now ask automated tools not to read them
What the site said: On four cards for candidates with no cannabis record, the list of what we checked named a page as read, and said what was on it, on a site whose robots.txt now names this guide’s crawler and asks it not to read the site: the candidate’s own campaign site on the cards of Odus E. Evbagharu (Texas House District 135) and Tim McDonough (Texas House District 114), and the Blue Voter Guide’s district page on the cards of Melanie Medley-Thomas (Texas House District 103) and Anissa Chilmeran (Texas House District 61). Each of those pages was described as carrying no cannabis content. The same kind of claim, about her own campaign site, sat in the search record this guide keeps for Jenn Mack, a certified write-in for Governor; write-ins are listed by name only and that record is not printed anywhere on the site, but it has been withdrawn the same way. The same withdrawal was made on Lauren B. Peña’s card (U.S. House District 37) on 2026-10-20 and was not recorded in this log at the time.
What it says now: Those claims are withdrawn and the pages are no longer listed as read. Each card says, in one sentence, what it used to claim and why it was withdrawn, and nothing from those pages is relied on. No position was added or removed and no score moved: none of these candidates has a cannabis record on this site, before or after. Lauren B. Peña’s withdrawal had been published as two sentences, and the second was left showing on her card with nothing before it to explain it; it is now one sentence, in the same words.
How the error was found: The robots check this guide runs before every build asks each site’s robots.txt again once our saved copy is more than a day old. On 2026-10-27 it found three of the cited addresses refused: jennmackfortexas.com and the two bluevoterguide.org district pages. On 2026-10-30 every page that any of the 178 no-record cards claims to have read, 677 claims in all, was asked about again, and two more campaign sites, odusfortexas.com and timforhouse.com, now refuse this guide’s crawler by name. The copies of these robots.txt files we held when the pages were read have been replaced by the current ones, so whether each site permitted us at the time can no longer be shown; the two Blue Voter Guide pages had answered normally when read on 2026-10-06. The claims are withdrawn either way.
Two cited pages had stopped working, and the cards did not say so; both now cite a working copy of the same page
What the site said: Alan Blaylock’s card (Texas House District 93) cited Community Impact’s candidate Q&A for his race at an address that answered HTTP 404. Sarah Eckhardt’s record of her campaign Issues page (Comptroller) cited that page on her own site, where it answered HTTP 404. Neither card said so. Sarah Eckhardt’s record also said the Internet Archive had returned no capture of the page, which was wrong: the Archive holds several.
How the error was found: The link check of 2026-10-27 recorded both addresses as answering HTTP 404 on two attempts. On 2026-10-30 each was asked again, after its site’s robots.txt was checked at that exact address, with one request and no redirect followed, and each still answered 404. Community Impact had changed the page’s address, and the new one answered 200 with the same title, questions and answers. Her site’s own sitemap no longer lists an Issues page; the Internet Archive’s index lists captures of it from 2013 to 2026-08-05, and the quoted passages were checked against the 2026-08-05 capture, verbatim.
One record said a legislator filed a hemp bill he joined later, and another counted one bill twice
What the site said: Jon Rosenthal’s page said he "Filed HB 2818 as author" on 2023-02-24, citing only a page of the Legislature’s bill-lookup site that this project is not permitted to read. James Talarico’s page carried two records of his authorship of HB 1831 in 2023: one dated 2023-02-03, the filing, and a second dated 2023-03-01, a date nothing in the House Journal supports. The second made his card count the same bill twice among the records behind his score.
What it says now: The House Journal prints HB 2818 at first reading as "(By Cain)" (journals.house.texas.gov/hjrnl/88r/html/88RDAY22FINAL.HTM) and, in its caption on 2023-05-01, lists the authors as "Cain, Rosenthal, T. King, Toth, Bernal, et al." (journals.house.texas.gov/hjrnl/88r/html/88RDAY52FINAL.HTM) Rosenthal’s record now says he was listed as a joint author on that date and cites the journal. The date he signed on is printed only on the bill-lookup page, so the record gives the journal’s date instead. Talarico’s duplicate record was removed; his score did not change (83), and his card now counts the bill once. Checked against the House Journal for 2023-03-14 and 2023-05-01.
The ten cards whose Houston voters' guide address stopped working now also cite a live League of Women Voters page for the same race
What the site said: On 12 October this site disclosed that, on ten no-record cards for Houston-area Texas House candidates, the League of Women Voters of Houston guide page each card cited had stopped working, and that we had kept no copy, so a reader could not check that finding. After that disclosure the cards still carried no League source a reader could open. One of the ten, Frank Salazar's, also said that whether this candidate had a campaign site of their own was an open question, because this guide had not found one.
What it says now: Each of the ten cards now also lists the League of Women Voters' VOTE411 page for the candidate's race in the November general election, read on 17 October, answering HTTP 200 and naming the candidate. The Houston chapter says on its own guide that the answers it collects are published on Vote411.org. It is the League's four-question general-election set, not the two-question primary set the dead pages carried. Kyle Harding, Alexandria Nicole Butler and Mike Michna have not responded on it; Heidi Hall answered two of the four questions; the other six answered all four. No answer from any of the ten mentions cannabis, hemp, THC or drugs, so no position was added to any card. The ten old addresses were fetched again on 17 October and every one still answers HTTP 404, so the disclosure on each card stays, now with that date. Frank Salazar's card now also lists a campaign site of the candidate's own, which the candidate's VOTE411 entry gives as the campaign website and which names the district; it carries no cannabis content, and the sentence saying no site had been found is replaced. Kyle Harding's card also lists Community Impact's District 139 Q&A, which records no response from the candidate. The chapter's printed general-election guide is online as a flipbook, but its pages are served from a host whose robots.txt this guide cannot obtain, so nothing in it was read and it is not cited.
Two more no-record cards now say which opening rolls of the Legislature were read for their candidate
What the site said: On the cards for David Flores (Texas House, District 144) and Eric Flores (U.S. House, District 34), this site still said it could not tell a reader whether the candidate had held a seat in an earlier Texas Legislature, because it had found no source it was permitted to read that carries the roll of former members. On 2026-10-16 these two cards were held back from the narrower sentence the other cards received, because the Senate’s first-day rolls of 2019 and 2025 print one senator by the surname Flores alone.
What it says now: Both cards now carry the same narrower sentence as the others: no one by the candidate’s name appears in the House Journal’s first-day rolls of members-elect for the 75th Legislature (1997) through the 89th (2025), or among the senators the Senate Journal names with a district on the first day of the 76th (1999) through the 89th, which is narrower than a negative answer. Each card also says who the surname-only senator is: Pete Flores, named in full by the Texas Senate’s own member page as the senator for District 24, with Senate service from 2018 to 2021 and from 2023; by the Secretary of State’s official results as Peter P. (Pete) Flores, winner of the 2018 special runoff election in Senate District 19; and by the Senate Journal on the first day of the 88th Legislature. senate.texas.gov/member.php?d=24senate.texas.gov/seniority.phpelections.sos.state.tx.us/elchist332_state.htmjournals.senate.texas.gov/SJRNL/88R/HTML/88RSJ01-10-F.HTM
How the error was found: The two first-day rolls were re-read to confirm that each prints exactly one senator by the surname Flores, beside full-name rows that together account for all 31 seats. The Senate’s member page and seniority list and the Secretary of State’s results page were read on 2026-10-17, each checked for permission first and each answering HTTP 200. Nothing was concluded from the shared surname alone: the change rests on a source naming that senator in full, and the name it gives is not either candidate’s.
Six cards still said a candidate had not completed Ballotpedia’s survey after they had
What the site said: Six no-record cards said Ballotpedia records that the candidate did not complete its Candidate Connection survey. That was what the profile showed when we read it on 2026-09-20, but the sentence was written in the present tense, and each of these candidates has since completed the 2026 survey.
What it says now: Each card now says that when we re-read the profile on 2026-10-15 it recorded a completed 2026 survey, that the answers contain no occurrence of cannabis, marijuana, marihuana, hemp, THC, CBD, delta-8 or Compassionate Use, and that on 2026-09-20 it had recorded none. The 73 other cards whose Ballotpedia survey statement still holds now give the date we confirmed it, so none of them reads as a present-tense claim.
How the error was found: Ballotpedia refused automated requests on 2026-10-12 and became readable again on 2026-10-15. Every cited profile was then re-read and compared with the copy saved on 2026-09-20.
Nine cards said a candidate had no Ballotpedia survey when Ballotpedia showed one
What the site said: Nine no-record cards said Ballotpedia records that the candidate did not complete its Candidate Connection survey, or that their profile carries no such survey. For each of them Ballotpedia showed a survey the candidate had completed for the 2026 election. For seven, the copy of the profile we read on 2026-09-20 already showed it. The other two cards also said, a few sentences later, that the candidate’s survey answers were published on that page, and Ballotpedia dates those answers to before our sentence was written. The cause was ours: the tool that summarised each profile looked for “did not complete” anywhere on the page, so a sentence about a candidate’s survey in an earlier election, 2020, 2022 or 2024, outweighed the sentence about the 2026 one.
What it says now: Each card now says when we re-read the profile (2026-10-15 or 2026-10-16), that it records a completed survey and in which year it was completed, and that the candidate’s answers contain no occurrence of cannabis, marijuana, marihuana, hemp, THC, CBD, delta-8 or Compassionate Use, together with a note that the earlier sentence was our error. None of these candidates gains a cannabis record, because none of the surveys raises the subject.
How the error was found: A re-read of every Ballotpedia profile this guide cites, 167 pages, all HTTP 200, was checked first against three candidates whose survey status was already known. One of them, a candidate whose survey this guide quotes, came back as not having completed it. That failure led to the faulty rule. Survey status is now read per election from the heading Ballotpedia prints above each survey.
180 no-record cards now say which opening rolls of the Legislature were read for their candidate
What the site said: On 180 candidate cards with no cannabis record, this site said it could not tell a reader whether the candidate had held a seat in an earlier Texas Legislature, because it had found no source it was permitted to read that carries the roll of former members. The question had been left open there since an earlier answer, drawn from a site that asks this one not to read it, was taken down on 2026-10-07.
How the error was found: Each chamber publishes its own journals, and the Legislature’s own journal index lists every first-day journal; both journal sites were checked for permission before anything was read, and only pages that came back whole were used. A House roll was accepted only when it accounted for District 1 through District 150 in order. Before any card was changed, the same name match was run on candidates already known to have served and found them where the journals put them — Ana Hernandez in House District 143 from the 80th Legislature through the 89th, Pat Fallon in House District 106 and then Senate District 30, Lois Kolkhorst in House District 13 and then Senate District 18 — so the empty result on these cards comes from a check that finds a former member when there is one.
Quotations in eight court-race notes and four candidate cards did not name the page they came from
What the site said: Eight court-race notes quoted court opinions, the Fifteenth Court of Appeals’ own page, the Supreme Court of Texas’s page on its justices and the Texas Tribune’s guide to the Fifteenth Court of Appeals races, and four candidate cards quoted the candidate’s own website, without printing the address of the page each quotation came from. The Court of Criminal Appeals Place 4 note named four opinions by case number but not where they were read. One card, Leticia Gutierrez’s, put in quotation marks a description of her that her website carries only in its page-sharing description, not in the text a visitor reads.
What it says now: Every one of those quotations now links the page it is taken from, and each was checked word for word against a copy of that page fetched only after the host’s robots.txt permitted it. None of the quoted words changed, because every one was found as printed. Leticia Gutierrez’s card now says, without quotation marks, that her site’s own page description names her as the Democratic candidate for the U.S. House of Representatives in Texas, District 9.
How the error was found: A new check, pipeline/verify_prose_quotes.mjs, compares every quotation in race notes, no-record cards and the other published prose against the stored copy of a page the same record cites. It found these quotations in no cited page and traced each one to the page it actually came from. It now runs on every build (tests/prose_quote_guard.mjs), so a quotation that names no page carrying its words stops the build.
Fifty-five cards described a search for the candidate’s own campaign website that this guide cannot show, and 46 of them concluded from it that no such website exists
What the site said: Fifty-five no-record cards carried an account of a search for a campaign site of the candidate’s own. It named two queries — the candidate’s name with the office or district and the words “campaign website 2026”, and the same with “issues platform priorities” — and said that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand. On 46 of the 55 it ended, in capitals, that no campaign site of their own had been found. That is the strongest kind of negative this site publishes about a named living person: not that we have nothing, but that we looked properly and there is nothing. No record of any of those searches existed. Not a query, not a result, not the name of a search engine, not a note of a failure, anywhere in this project, while every other kind of evidence on the same cards had one.
What it says now: The account of the search is gone from all 55 cards and the conclusion is withdrawn. In its place each card says that whether this candidate has a campaign site of their own is an open question here and not a negative answer: this guide has not found one, and no longer claims to have looked in a way that could settle it — so if the candidate publishes a platform of their own, we may simply not have reached it, and the absence of a cannabis position on the card is weaker evidence than it looks. The nine cards where a site had already been found by another route keep everything they say about that site; only the account of the searching is withdrawn from them. Nothing has been put on any card claiming that a site of their own does or does not exist.
How the error was found: The absence was noticed first: every other evidence class on these cards has an artifact and this one had none. Before deciding, both queries were run again on 2026-10-12 against a known positive — a candidate whose own campaign site this guide has already read and published a record from, and whose campaign has a page headed "Legalize Cannabis". Nineteen results came back across the two searches. Every one was a voter-guide aggregator or a news report, which are the two classes the withdrawn sentence says it discarded before concluding, and none was on her own domain; her site was confirmed live the same day, after its robots.txt was checked and found to permit it. So the method writes down "no campaign site of their own was found" about people who have one, and the searches were not re-run, because a re-run would have reproduced the same false negatives with a log attached and that would have looked like proof. The queries, every result, the gate verdict and the test are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json, and tests/search_negatives_have_evidence.mjs now refuses to let this site publish a search that came back empty without naming something a reader can check.
A notice on 40 cards said no campaign website for the candidate could be located, when some of those websites had been located and one had merely not been read
What the site said: Cards with no source of the candidate’s own carried a notice in bold reading "No campaign website for this candidate could be located or loaded." On every card that also carried the withdrawn search above, that notice and that search were the same assertion: the search was the proof and the notice was the verdict. It was also wrong for a second group for a different reason — several of those cards name the candidate’s own website by address, as a page we could not read because its host asks automated tools not to. Those websites were located, precisely and by address. Only the reading did not happen.
What it says now: The notice reads "No source of this candidate’s own is on this card." That is a statement about this page rather than about the world, and a reader can check it against the list of sources printed directly beneath it. The rest of the notice is unchanged and slightly stronger: everything below was published by someone else, so the absence of a cannabis position is weaker evidence than it looks — it may mean only that the candidate has not published a platform anywhere we have found, or has published one somewhere we are not permitted to read.
How the error was found: It followed from withdrawing the search above. Deleting the evidence and leaving the conclusion standing in bold would have been the quiet edit this site promises not to make, in its worst form.
Nineteen cards said in their search log that we had read the candidate’s own campaign website, and then left it out of the list of sources checked
What the site said: Nineteen no-record cards name the candidate’s own campaign website by address, in their search log, and say it was read in full — most of them with the number of pages and the number of characters. None of the nineteen listed that website in the panel headed "Campaign sites and voter guides checked", which is the part of the card a reader opens to see what was looked at. The panel listed the League of Women Voters questionnaire and the newsroom profile and omitted the candidate’s own site. The log and the panel disagreed about whether we had read it.
What it says now: All nineteen list the site in the panel, marked as the candidate’s own, with the page and character counts the log already published. No new page was fetched and no new claim was made: each entry is built out of the sentence the card was already publishing.
How the error was found: Comparing what the cards say in prose against what they record as structured sources. Every one of the nineteen addresses was put to the robots gate again, at the path cited, before being written down, and all nineteen are permitted. Two further cards look like these and are not: they name a site in a sentence that withdraws an earlier claim to have read it, and they were left alone.
Ten cards cited a page that no longer exists, as evidence of how carefully we had looked
What the site said: Ten Texas House candidates have a card saying we found no cannabis record for them. The whole purpose of that card is to show a reader that the blank is the record and not our laziness, so each one lists what we read. On these ten, one of the sources listed was the candidate’s entry in a Houston voters’ guide, described as read in full by us on 6 October. Every one of those addresses now returns "not found". The guide is still publishing and still permits automated reading, so this is not a host shutting us out — the pages were taken down, and that guide’s own candidate index no longer carries these names. We did not keep a copy. So the single most load-bearing claim on the card, that we checked, was pointing at nothing a reader could open, and nothing we could re-check either.
What it says now: Each of the ten citations now says plainly that the address no longer works, that the page was there when we read it, that we kept no copy, and that it is therefore weaker than every other source on the card. We did not delete the findings. Each describes the page in enough detail — one names its heading and its two-question structure — that they were read as described, and deleting accurate work is its own error. What a reader was owed is the fact that they cannot check it, which is now on the card. Checked by fetching each address again on 12 October, by confirming the guide’s own index no longer lists these names, and by confirming the host still permits us, so the failure is the page and not a refusal. This is the second time the site has published a citation to a page that was not there; the first was on a different host, and the check that caught it had never been widened beyond that one host.
One US House candidate’s score counted a vote he did not cast and missed one he did
What the site said: Colin Allred, the Democratic candidate in US House District 33, published a score of 84. Two of the records behind it described more than one vote each, and our scoring could only read one vote per record. The first read "Recorded Yea on House Roll Calls 103, 104 and 105, three amendments to H.R. 3617, the MORE Act", and because it named the MORE Act our scoring counted it as a Yea on the MORE Act itself — the strongest supportive federal vote this site recognises — when it was a vote on three amendments to it, which this site scores at one point, nothing and nothing. The second read "Recorded Nay on House Roll Call 106, the motion to recommit H.R. 3617, the MORE Act, and Yea on House Roll Call 107, its passage, the same day". Our scoring read the first half, saw a procedural motion, and set the whole record aside — so his Yea on passage of the MORE Act, which he did cast, counted nowhere. The eighteen other Texans who hold those same amendment roll calls each hold them as separate records and were scored correctly throughout.
What it says now: His five recorded votes of 2022-04-01 are five separate records, as the Clerk of the House records them: Yea on roll call 103 and Yea on roll call 104, the two study amendments, both shown and neither scored; Yea on roll call 105, the security-clearance amendment, scored at one point; Nay on roll call 106, the motion to recommit, shown and not scored because a procedural motion is not a position on a bill; and Yea on roll call 107, passage of the MORE Act, scored at two points. His published score is 83, from ten scored records instead of nine. No other candidate’s score, band or confidence word changed.
How the error was found: Both halves were fixed in one change, because they moved the same number in opposite directions and fixing either alone would have published a number that was wrong by the other. Every one of the five votes was read back from the Clerk of the House’s own roll-call record for that roll, and each roll was checked against the tally the Clerk prints in the same file before any vote was attributed. The scoring rule that should have refused the first record was also repaired: it was written in the singular and the word "amendments" walked past it.
One committee decision moved ten legislators’ scores twice
What the site said: Ten members of the Texas House sat on the House Committee on State Affairs when it handled Senate Bill 3, the intoxicating-hemp-THC ban, in 2025. The committee reported the substituted bill to the full House on 30 April, and on 5 May it put the identical question a second time, on a motion to reconsider, and reported the same report again. This site counted the two record votes as two separate positions on the bill, each at the full weight of a committee vote, so one committee decision moved ten named people’s published numbers twice. For four of them the second copy was decisive in a particular way. Senfronia Thompson, Rafael Anchía, Yvonne Davis and Richard Pena Raymond are each recorded Nay twice on the House floor on the bill and Aye twice in committee to report it, so their Senate Bill 3 record netted exactly zero out of 24 weight, and the defining cannabis bill of the 2025 session contributed nothing at all to their score.
What it says now: The 5 May record vote is still printed in full on all ten pages, with a note beside it saying that it is the same committee report already counted and that the number rests on the 30 April report alone. Ten published scores moved, every one of them upward, none across a band boundary and none changing its confidence label: Lacey Hull 53 to 56, Senfronia Thompson 68 to 72, John W. McQueeney 34 to 38, Charlie Geren 58 to 61, Rafael Anchía 71 to 75, Yvonne Davis 68 to 72, Richard Pena Raymond 68 to 72, Will Metcalf 27 to 29, Drew Darby 56 to 60 and Ken King 44 to 46 — six Republicans and four Democrats. Checked by recomputing all 436 candidates’ scores before and after: those ten are the only numbers that moved.
How the error was found: Two things settled it rather than one. This site had already ruled, in the correction published about a reason for vote filed on Record Vote 3192, that a document explaining a decision already counted is one decision counted twice; and the methodology page already told readers that where a question is put again on a motion to reconsider, only the vote that stood counts, which is not what the arithmetic was doing. What proves these are one decision and not two is that it does not matter which copy is set aside. Recomputing all 436 candidates with the 5 May copy removed, and again with the 30 April copy removed, moves the same ten numbers by the same amounts, because the two are committee votes of the same weight, the same direction and the same recency band. Two records that are interchangeable are not two positions. The error itself had been sitting in an audit file that the build does not run, and that file stopped at its first failing check, so it reported one defect and went quiet about this one. It is now checked by tests/score_document_identity.mjs, which every build has to pass.
Our methodology page said an old record is weighted down. For 116 of the 253 published numbers it is not weighted down at all
What the site said: The methodology page set out the weight given to each kind of act and then added, in the same sentence, that an act is weighted "by recency: full weight within two years, 0.6 out to six years, 0.3 beyond. Positions change, but an old vote never stops counting." A reader takes that to mean an older record counts for less. It does not, except against a newer record belonging to the same candidate. The recency figure multiplies the top and the bottom of the same weighted average, so wherever all of a candidate’s scored records fall in one age band it cancels out exactly. Re-dating every record on this site to the day before this entry was written leaves 116 of the 253 published numbers unchanged by a single point, and where a number rests on one record it cannot move at all. The page printed the contradiction beside itself: the Railroad Commissioner page described a 2014 campaign plank, written for a run at a different office, as "weighted for its age" directly above a score of 100 out of 100, the top of the scale, with the written reason saying the largest available discount had been applied.
What it says now: The page now says what the figure does. It states that recency weighs a candidate’s records against each other and does nothing else, that the factor cancels exactly where every record sits in one age band, and that a twelve-year-old statement on its own therefore produces exactly the number the same statement made yesterday would. The three counts in that paragraph — the numbers that do not move, the candidates holding a record more than two years old, and how many of those are unaffected — are recounted from the data by pipeline/make_methodology.mjs on every build, so none of them can go stale, and the generator refuses to build the page if the count ever reaches zero without the paragraph being rewritten. The written reason on the Railroad Commissioner record no longer claims an age discount; it says instead that this is the only record on file, so there is nothing newer for it to be weighed against. tests/recency_claim.mjs fails the build if the retracted sentence returns, if the page stops stating what the factor does, if the printed figures disagree with a live recount, or if any written reason again tells a reader a record is weighted for its age. No published score changed: the arithmetic was not touched, and the error was in what the page said about it.
Thirteen quoted questionnaire answers carried a second citation to a newspaper’s index for one day
What the site said: Thirteen Texas House, Texas Senate and US House candidates are quoted here answering the Fort Worth Report’s 2026 candidate questionnaire. Each quoted answer cited the candidate’s own questionnaire article, and then cited a second address, fortworthreport.org/2026/02/16, beside it as though it were further evidence. That address is the newsroom’s archive index for a single day. It carries none of the 26 quoted sentences, it names no candidate’s answer, and a reader who followed it to check what we had written arrived at a list.
What it says now: The date index is gone from all thirteen. Each quoted answer cites the candidate’s own questionnaire article and nothing else: that is where the answer is published, and it is the copy every one of those quotations was checked against, word for word, before and after this change. No quotation changed and no candidate’s record gained or lost an act.
How the error was found: A sweep for citations a reader cannot follow. The index was found by asking, of every source listed under a quoted act, whether our copy of it contains any of that act’s quoted sentences. A source we have read that carries none of them is either corroborating something other than words — a roll call, a bill history — or is not the document at all, and this one was not the document.
Two cards said a candidate’s own campaign site could not be read, after it could
What the site said: Leticia Gutierrez’s card said her campaign site “WAS NOT RETRIEVABLE” and Caitlin Rourk’s said the same of her issues page. Both sentences described September, when those hosts published a robots.txt that named our readers one by one and told each of them to read nothing. Those files changed on 2026-09-20 and both sites were read in full on 2026-10-06, Rourk’s issues page among them. Two further cards, Claire Reynolds’s and Chris Jimenez’s, offered a partial page count — 25 pages of 51, 13 of 16 — as the measure of how thoroughly we had looked, when by then every page had been read.
What it says now: All four cards now say what each host asked for, when it stopped asking it, and that every page in the site’s own sitemap was read on 2026-10-06 — 15, 51, 16 and 12 pages — together with the site’s own full-text export where it publishes one, and that a search on word boundaries for eleven cannabis terms over the extracted text finds none of them anywhere on any of the four sites. Nothing was added to any of the four records and nothing taken away: all four sites are silent on cannabis, which was the finding before, and it now rests on the whole site instead of part of it.
How the error was found: Finishing a sweep of hosts that had reopened to us. A sentence saying a source could not be read is a claim with a date on it, and it has to be asked again when the reason it was true stops being true.
Our Texas law record stopped on 31 July and said an executive order had done the work
What the site said: The Texas row on the front page and on /explainer/cannabis-law-by-state/ ended its account of hemp-derived THC with one clause: SB 3 was vetoed on 2025-06-22 and “comparable restrictions were then imposed by executive order and health-department rulemaking, effective 2026-07-31.” Two things were wrong with that and one thing was missing. What took effect on 2026-07-31 was a Department of State Health Services scheduling notice, not an executive order; the governor’s 2025 order is a separate act a year earlier and we could not read it at a permitted source, so it is no longer cited. And the restrictions are not comparable to SB 3: the notice puts manufactured delta-8 and the other tetrahydrocannabinols back into Schedule I, while the department’s own page says a consumable hemp product may still contain delta-9 THC at not more than 0.3 percent on a dry weight basis. Missing entirely was the litigation. The row carried no trace of the Supreme Court of Texas case the notice was issued under, of the Travis County injunction over how that 0.3 percent is measured, of the federal suit in Galveston, or of the temporary-injunction hearing set for 2026-10-30, which falls inside early voting.
What it says now: The row carries the record with dates and citations a reader can open: No. 23-0887, Texas Department of State Health Services v. Sky Marketing Corp., decided 2026-05-01, quoted from the court’s own opinion PDF; the 2026-05-15 stay order and the 2026-06-05 denial of rehearing, read off the court’s orders pages; the reinstatement notice at 51 TexReg 4597, published 2026-07-10 and effective 2026-07-31; the signed Travis County temporary injunction of 2026-05-01 in Cause No. D-1-GN-26-002511 and the State’s pending appeal No. 15-26-00099-CV, under which that injunction is not in force; and the two undecided matters, the federal restraining order refused in August and the Travis County hearing set for 2026-10-30.
How the error was found: Every source was put through the robots gate at the path cited and fetched only on allow. The court’s opinion, its orders pages, the Texas Register notice and the health department’s pages were all read in full. What could not be read is named in the row itself rather than smoothed over: search.txcourts.gov refuses us, so the Fifteenth Court of Appeals docket, the State’s reply brief and the court’s own notice of the 2026-12-08 argument are cited only as the parties describe them; the federal docket is on PACER, so the Galveston order is attributed to the Texas Tribune’s report and not stated as a record; and the date on which the second Travis County restraining order was refused is left undated, because the report does not give it and the docket that would is closed to us.
Our methodology page left out twenty-nine of the rulings it said it contained
What the site said: /methodology opens by calling itself “the whole method, including the judgments”, and every candidate card links to it as the place to go and find the exact line to disagree with. It published the Texas bill table and the Texas bills we deliberately do not score, and then stopped. The twenty-nine federal rulings in the same file were on none of it: the MORE Act in four Congresses, both House passages of the SAFE Banking Act, six floor amendments that carry their own direction, and seven deliberate exclusions — among them an omnibus funding vote left out because it separated nobody but party, four appropriations roll calls left out after measuring that the same two members cross on bills whose cannabis riders point in opposite directions, and a bill-number collision in which a vote on a tax act was counting as cannabis support until the Congress was added to the key. Those twenty-nine are what every number on a US House candidate’s page rests on, so the page that exists to be audited showed the method behind the Texas scores and none of the method behind the federal ones. Two smaller claims on the same page were also wrong: it said four federal amendments carry their own direction when the table carried six, and it listed three kinds of act it leaves out of the arithmetic where the code leaves out more.
What it says now: All 164 rulings in the file are on the page. The federal ones are rendered in the same shape as the Texas ones — a table of what a Yes vote counts as, with the subject, the reason and a link to the Clerk of the US House or to congress.gov for each row, and a separate list of the seven votes we deliberately do not score, each with the argument for leaving it out in full. Every federal count in the prose is now read from the table rather than typed, so the “four amendments” sentence cannot drift again. The page also now describes four things the arithmetic skips that it had never mentioned: a record we read no direction in at all, which was 782 of the 4,501 records on this site across 163 candidates when this entry was written, every one of them published with our written reason for it — the page recounts both figures from the data each time it is built, so neither is a number typed into prose; a vote whose record does not say which way the member voted; a motion to reconsider, where only the vote that stood counts; and the federal funding votes. It says which recorded votes do count — final passage, a second or third reading, a motion to concur, a conference committee report, a committee vote to report a bill favourably, and in the US House a motion to suspend the rules and pass, which is final passage there even though suspending a reading rule in Austin is procedural — and which session or Congress a bill number is matched on. No published number changed: the scoring code was not touched, and the error was in what the page said about it.
How the error was found: The gap was found by reading what pipeline/score.mjs actually reads and comparing it, line by line, against what the page describes, rather than by reading the page on its own. tests/methodology_rulings.mjs now fails the build if any ruling in the file — Texas or federal, scored or deliberately not scored — is missing its citation, its subject, its source or its reason from the published page, and it compares the reason as a reader sees it, after the typography pass, so a row cannot pass by being half there. It was proven to fail by dropping one row from the generator and rebuilding: it named the missing ruling, and it named the amendment count when that sentence was typed back to “four”.
With JavaScript off, neither page said the registration deadline had passed
What the site said: The front page and /vote both led, bolded and first, with "Registration deadline: Monday 5 October 2026." That is a statement of the calendar and it was never false. It was also spent: registration had closed two days earlier, and the thing a reader could still act on — early voting, opening 19 October — was unbolded and second. With JavaScript turned off, which is also what a search engine and an archive keep, nothing on either page said registration had closed at all: the words existed only in a script-driven line and in a card two-thirds down /vote, conditioned on "If the lookup does not find you". The hero counted 27 days to election day and never mentioned the sooner date, so the one number at the top of the site was the later and less useful of the two.
What it says now: Both pages are built from one card that splits past from future at the build date. A deadline still ahead keeps the never-false calendar label it always had; a deadline the build date has already passed is printed in the past tense, from wording the data file has carried for all five dates since it was written — so with no JavaScript at all the page now reads "The deadline to register for this election was Monday 5 October 2026." above nothing, and the early-voting window leads in bold. ⚠️A date is never pre-dated into the past: it moves tense only once the build is strictly past it, which makes every sentence monotone — true on the day it is built and on every day after — so a stale build can only under-report, which is exactly what the old card did. The front-page countdown now counts to the soonest deadline still ahead and names it, instead of to a date typed into the hero. The card stamps the date it was built on, tests/vote_degrade.mjs cross-checks that stamp against the one the status line is written from by a different generator in the same build, and tests/vote_degrade.mjs, tests/vote_passed_deadline.mjs and tests/hero.mjs each execute the published script across the whole election window and are re-run against deliberately broken copies of it that they must reject. No date changed.
We read 1,091 pages from a library that had already refused us by name, and put the search on 318 pages as proof of our thoroughness
What the site said: Three hundred and eighteen pages of this site carried a sentence of this shape: "Also checked on 2026-09-23: the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846, searched by surname at lrl.texas.gov and every matching member’s profile read for the sessions they served — no member who served in the 84th Legislature (2015) or later shares this candidate’s surname and first name, so there is no earlier Texas legislative voting record to report." It stood on the no-record cards of 174 candidates, and ten scored acts belonging to three sitting members cited the same host to say which of two same-named legislators had cast a vote. lrl.texas.gov publishes a robots.txt that says "User-agent: ClaudeBot / Disallow: /", and names anthropic-ai and Claude-Web the same way. Our own copy of that file was saved to our own disk at 12:42 on 2026-09-22. At 06:55 the next morning, eighteen hours later, we began fetching, and we took 1,091 pages. Our robots gate cleared every one of them because it read the rules the site publishes for everyone else and ignored the group that names us; the comment in our own code asserted that we are "not a named crawler", and that was simply untrue. So the single sentence on this site that was meant to show a reader how far we had gone to check a candidate was the one we obtained by going where we had been asked not to go.
What it says now: The sentence is withdrawn from all 174 cards. In its place each card says that we read the Library, that the Library had asked automated tools not to, that the refusal was already on our disk before we fetched, and — the part that costs us something — that the card therefore no longer tells the reader whether this candidate served in an earlier Legislature. That is now an open question and not a negative answer: if one of those 174 held a seat before the 89th Legislature, there may be a cannabis voting record this site is missing, and we have not found a source we are permitted to read that carries the same roll. We considered keeping the sentence on the ground that it is factually true — we did read the Library, so the search did happen. We withdrew it anyway, for the same reason we withdrew a published zero out of a hundred and a set of quotations from a state website this week: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one is evidence that it could not be. That it is true is what makes it worth withdrawing. The ten scored acts keep their votes, on better evidence than They had. Four acts of Claudia Ordaz’s rested on the Library to connect the journal’s "Ordaz Perez" to the name on our card. The House Journal, which permits us, says it outright: "Claudia Ordaz Perez, District 79, requested to have her name displayed as Claudia Ordaz." The Library was never needed, and the entries now cite the journal alone. The other six, three for Caroline Harris Davila and three for Cody Harris, turned on expanding the journal’s "Harris, C.E." and "Harris, C.J." into full names, and we had taken those full names from the Library. ⚠️We have no permitted source that expands either initial, so which Harris is which is no longer established here and the entries say so. It changes neither record: we went back to the journal’s own rolls for all three votes, and on every one of them Harris, C.E. and Harris, C.J. are printed in the same list — both Yea on Record Vote 143, both Yea on Record Vote 153, both Nay on Record Vote 605 — so the vote shown is the vote either member cast. ⚠️The 1,091 pages have not been deleted. Deleting them would not unmake the request and would destroy the only record that we made it. They are set aside where no code names them, the gate now refuses every path on that host, and the sweep that fetched them exits rather than running. The gate also used to be consulted only before a network call, so a page already on our disk from a refusing host was handed straight back; it is now consulted before the cache as well, because reading a refusing site out of our own cache is the same act as reading it through an archived mirror, which this site has already had to withdraw quotations for.
How the error was found: Not by a test, and not by anyone looking at the pages. It was found by reading the timestamps on our own disk next to each other: the robots file we had saved for that host, and the pages we had saved from it. The refusal was the older of the two. Nothing in the pipeline had compared those two dates, because the gate had already returned "allowed" and nothing re-asks a question it believes it has answered. ⚠️The same reading found a second case we have not yet resolved, and it is larger. capitol.texas.gov asks every automated tool to stay out of /BillLookup/, and 2,779 of this site’s 4,378 scored acts cite a page under that path — 438 of them cite nothing else. Unlike the Library, that host had never been asked at all when we read it on 2026-09-18; we only saved its robots.txt on 2026-10-02. The gate now refuses those paths, so six of this project’s research tools can no longer fetch them, which is the gate working and not a fault. What to do about the 438 acts that rest on them is a decision this entry does not pretend to have made.
Three candidates' names were printed without the accents they carry
What the site said: The Senate District 26 page named the two candidates in that race “Jose Menendez” and “Julian Villarreal”, and the House District 103 page named its member “Rafael Anchia”. All three carry an acute accent: José Menéndez, Julián Villarreal, Rafael Anchía. Dropping an accent from a name is not a typographic choice, it is a different spelling of a person, and these three were the only candidates on the site whose names were spelt without one.
What it says now: The three names are printed with their accents. Each was checked against the source closest to the person. Julián Villarreal spells his own first name with the acute in all seven places it appears on his campaign site, including the “Pol. adv. paid by” political-advertising line, and the Green Party of Texas heads its page for him the same way; the Secretary of State's certified ballot also prints JULIÁN Villarreal, so every source agreed and only this site did not. The Texas Senate heads its own District 26 page “Senator José Menéndez” and addresses him “The Honorable José Menéndez”. The Texas House prints “Rep. Anchía, Rafael” in the title, heading and navigation of its District 103 page. ⚠️The certified ballot is not the authority on accents and is not being treated as one here: it prints jose Menendez and Rafael Anchia, but it is set entirely in capitals and carries a diacritic just four times in 1,395 pages, so its silence says nothing about a name while each chamber's own page for its own member does. No web address changed: the site builds a candidate's URL by removing accents, so all three pages stand where they stood.
How the error was found: By asking what mechanism could have removed one acute, and then testing the answer rather than assuming it. This project has a documented defect in which an ASCII-only character class deletes the letter underneath an accent, turning Peña into Pea. If that had run over the race file it could not have damaged one name only, so the prediction was that the other three accented names on the certified ballot would be damaged too, and the falsifier was finding them intact. They were intact: Lauren B. Peña, Marlón Durán and Sergio Muñoz, Jr. all keep their accents, and across all 436 candidates the deleted-letter signature appears zero times. Julián Villarreal had the letter present and merely unaccented, which is the opposite signature: a hand typing from an all-capitals list, not a program. The line that introduced it was then read, and both unaccented names in that race sit twelve lines apart in the same hand-written file. The sweep that followed compared every name the site prints against every accented spelling of the same name anywhere in its own data, which is what surfaced the other two; it also cleared a third, Ryan Guillen, whose name an outside organisation accents but whose own chamber does not.
House Bill 4918 now counts towards scores, and we did not say so
What the site said: An entry published on 5 October said of House Bill 4918 and House Bill 513 that "Both bills are still outside the arithmetic and no published number changed", and that whether HB 4918 should be scored was a question we had not answered. Since then we answered it and began counting the bill, and that earlier sentence stayed on this page saying the opposite. Scores moved with no entry here explaining why. This guide promises that corrections are never made silently, and this was a change made silently.
What it says now: House Bill 4918 of the 88th Legislature is scored, and a recorded Yea on it counts as support. We read the engrossed text: its three operative provisions strike the word "not" from Agriculture Code 122.301(b), repeal Health and Safety Code 443.204(4), and repeal the definition that provision relied on. It repeals a ban on one form of hemp product, so it widens legal access rather than narrowing it. 105 recorded votes across 99 candidates now count towards a number that did not count them before. ⚠️Only the final roll call counts, and that matters here. The House rejected the bill on Record Vote 1421 by 69 to 75, then voted to reconsider and passed it on Record Vote 1477 by 84 to 59. Counting both would have recorded a member who voted to pass the bill as neutral or worse on it, because the two cancel and drag the average towards the middle. The annulled vote is still shown on every page that carries it, with a note saying the House put the same question twice and the second answer is the one that stood. The 5 October entry below is left as it was written, because this log records what we said at the time; this entry supersedes it.
How the error was found: A check of whether any figure quoted in our outreach letters had gone stale. It had not — but the check turned up a ruling that changed 99 candidates’ pages with no entry on this page at all, which is a worse failure than a stale figure.
Thirty-seven US House race pages described no territory and did not say which map they were drawn on
What the site said: All 150 Texas House race pages open with the ground the seat covers — the District 20 page begins “Part of Cedar Park and Williamson County.” The 38 US House pages did not. Thirty-seven of them opened with the sentence “A United States House district.” and then went straight to the records, naming no counties and no map. Those 38 seats are the only districts Texas redrew for this election, so the one place a reader cannot assume the lines are the ones they remember was the one place the territory was missing. The effect was sharpest on exactly the journey this site is built to serve: the district finder tells a Cedar Park voter that they are represented by district 10 today and that the seat on their 3 November ballot is 17, and the district 17 page one click later described no ground at all.
What it says now: Each of the 37 opens with the seat’s territory and the map it is drawn on, in the style of the one blurb that already did — “A US House seat on the 120th Congress map, covering all of Bosque, Falls, Freestone, Hill, Limestone, McLennan, Milam and Robertson counties, part of south-east Williamson County, part of east Bell County and part of south Johnson County.” ⚠️Every county name and every compass word is measured, not remembered. The lists come from cross-tabulating the Census TIGERweb congressional boundaries against the Census county boundaries on one shared equal-area grid; the measurement, including the share of every county in every district, is published in the repository as pipeline/data/us-house-territory-2026.json. ⚠️The shares are of area, not of population, and the wording stays inside what was measured: “covering part of south-east Montgomery County” is a statement about ground, and a district can hold a third of a county’s land and most of the people who live there.
How the error was found: By following the site’s own best claim one click further. The finder’s redistricting callout is the sharpest thing on the site; the page it sends the reader to opened with a sentence that could have been printed under any of the 38 seats, and under it stood a member labelled as the incumbent of a district that no longer means what it meant. The counting was then mechanical: 37 of 38 US House blurbs began with the identical generic sentence, and 0 of 150 Texas House blurbs did. A test now fails the build if any US House blurb ships without naming the map it is drawn on.
An “Incumbent” badge on a redrawn seat claimed more than it could support
What the site said: Each of the 38 US House race pages badged the sitting member of that seat number “Incumbent”, in the same words the Texas House and Senate pages use. On a mid-decade redraw those are different statements. Pete Sessions holds seat 17 and has never represented Cedar Park; a Cedar Park reader who had just been told by this site’s own finder that their district number changed then met him, on the page the finder had sent them to, under a label that reads as “this is your current representative”.
What it says now: On the 38 congressional seats the badge reads “Holds seat 17 today” — the seat number, which is what incumbency on a redrawn map actually means — and a note beside it gives the map and the measured movement: how much of the district on the ballot was inside the same-numbered district before the redraw, where the rest came from, and that the figures are shares of area rather than of population. ⚠️The wording had to be true of all 38, including the seats that barely moved. District 13 kept 100.0% of its own territory and district 19 kept 100.0% of its, so a flat “this person did not represent you” would have been a second false statement in place of the first. Which seats moved is measured per seat and carried as data, so the renderer never guesses: districts 9, 35 and 32 kept under 3% of their old ground, and their note says so. The Texas House and Senate seats were not redrawn for this election and keep the plain badge.
How the error was found: Reading the race page as a voter arriving from the finder, rather than as a page about a seat. The badge was correct about the seat and wrong about the reader, which is why nothing on the page contradicted it. The amount each district moved was then measured rather than assumed, because the obvious fix — telling every reader the sitting member is not theirs — would have been false for five seats.
The zip-code lookup never mentioned the redraw, on the route the front page recommends
What the site said: “A zip code gives a quick answer”, the finder says. It was also the one route through this site that said nothing at all about the mid-decade redraw. The server asked its table for every map vintage and then dropped the current-map rows before answering, so the comparison the whole site exists to draw could not be made on that path. For ZIP 78613 — Cedar Park, the worked example this guide is built around — the answer correctly listed four ballot districts with their shares: 17 at 88.4%, 11 at 11.4%, 10 at 0.1%, and 31 clipping the edge. A reader who had been told elsewhere that they are in district 10 saw “District 10 — 0.1%” among three numbers they did not recognise, with nothing saying that 10 is their district today.
What it says now: A zip lookup now carries the same comparison the address lookup does, in the shape a zip needs: two lists rather than one arrow. For 78613 it says that the zip overlaps districts 10, 31 and 37 on the map that runs to the end of 2026 and districts 10, 11, 17 and 31 on the map used for the 3 November ballot, that the list shown is the ballot one, that numbers 10 and 31 appear on both lists without being the same ground on both maps, and that only a street address narrows it to one. ⚠️And the line that used to follow it was made unreachable when false. “Your congressional district number did not change” was printed whenever any current district was known and no single-number change had been worked out, which for a zip spanning three districts today and four in November would have been this site asserting the opposite of its own headline. It is now said only when there is one district on each map and they are the same one.
How the error was found: Reading the zip branch of the lookup against the address branch, line by line, after the race-page defect showed that the redistricting claim was being made in one place and dropped everywhere else. The filter that removed the rows sat four lines above the response that needed them.
The District 2 page said the seat covers Montgomery County; it covers part of it
What the site said: The one US House blurb that had been written by hand read “covering Montgomery County and part of north Harris County”. Montgomery County is split between two congressional districts on the map in force for this election: District 8 holds about 70% of its area and District 2 about 30%, the south-eastern part. The unqualified half of that sentence was the half nobody checked, and it was the model the other 37 were about to be written from.
What it says now: It reads “covering part of south-east Montgomery County and part of north Harris County”. ⚠️This is a statement about ground, not about people. The 30% is a share of the county’s area, and the county’s population is not spread evenly across it, so the figure must not be read as a share of Montgomery County’s voters.
How the error was found: The hand-written blurb was used as the test of the measurement, not the other way round. Its other clause — “part of north Harris County” — was reproduced from the geometry alone, which is the check that let the measurement be trusted for the other 37 districts; the same run put District 2’s share of Montgomery County at 30.1%, and the two halves of one sentence disagreed about how precise it was being.
Four citations to a public radio newsroom landed a reader on its front page, not on the article
What the site said: The Dallas and Denton entries in the local-law table, and the Houston charter-amendment entry, cited three KERA News articles by their keranews.org addresses. Those addresses no longer serve the articles: each one redirects to the station’s newsroom index at kera.org/news. A reader who clicked a citation to check what we said arrived at a list of that day’s headlines with no way to find the piece being cited.
What it says now: Those four citations are gone. Each of the three articles was already cited a second time as a dated Internet Archive capture of itself, and all three captures were opened and read today: they carry the articles in full, under their own headlines — "Dallas can decriminalize small amounts of marijuana for now, judge says", "Denton City Council repeals marijuana ordinance in split vote" and "Dallas agrees to block marijuana decriminalization ordinance after appeals court ruling". Those captures are now the citation, so every link under these entries opens the document it is offered as evidence of. Nothing we said about any of the three cities has changed.
How the error was found: A sweep for citations a reader cannot follow, which followed every cited address to its final destination rather than only checking that something answered. A redirect to a front page answers with HTTP 200, so a link checker that stops at the status code calls it healthy.
A candidate’s questionnaire answer was cited to a JavaScript file
What the site said: The Commissioner of Agriculture page quoted Clayton Tucker’s answer to The Dallas Morning News’s 2026 voter-guide questionnaire. Its first listed source was an archived copy of a five-megabyte JavaScript bundle from inside that guide, and its second was the guide’s own front page. The bundle does contain the answer — that is how we verified the quotation, and the wording is unchanged — but it is a program file, not a document, and no reader could reasonably be asked to check a quotation in it.
What it says now: The citation leads with the guide’s own page for that race and party, which is where the answer is published for a reader. The page was confirmed from the Internet Archive’s record of what it has captured, and the capture was opened: it is headed "The Dallas Morning News Candidate Q&A: Agricultural Commissioner - Democrat" and names him. The archived bundle is kept as the second source, because it is the copy the quotation was checked against. The guide’s front page, which pointed a reader at no particular race, has been dropped.
How the error was found: The same sweep. The bundle was flagged because its stored copy is five million characters of program text, which is not the shape of a document anyone cites.
The races index said it had been checked on a date before most of it existed
What the site said: The races section on the front page ended with the line “Checked against sources on 2026-09-18”, and every entry in our sitemap gave 2026-09-18 as the date the page last changed. The data behind that line had grown by 3,254 dated records since 2026-09-18, and carried acts dated as recently as 2026-10-06. So a reader was told the guide had been checked before most of what they were reading had been put into it — which is the one thing a freshness date is for.
What it says now: The line reads “Checked against sources on 2026-10-06”, the date of the most recent research pass, and the sitemap gives the same date. The date is a summary and not the authority: every act on every page carries its own date and its own sources, and every “no record” card dates each source it read, so a reader can always see when a particular thing was checked rather than relying on one date for the whole guide. ⚠️This is not a claim that all 4,377 records were re-read on 2026-10-06. No pass of that kind has ever been done, the data file now says so in its own notes, and the separate manifest in which a person records having verified new material has deliberately not been re-signed. The 38 records dated after 2026-09-18 are listed for that review.
How the error was found: Our own as-of test had been reporting it on every build for eighteen days — “carries dated material as recent as 2026-10-06 but claims it was checked on 2026-09-18” — and nobody could see it, because the build ran every test with its output thrown away and printed only pass or fail. The test had even written that limitation into its own notes. The build now replays the warning blocks of tests that pass, and a new test fails the build outright when a data file's date is older than the newest dated record inside that same file.
A member’s written reason for his own vote was counted a second time beside the vote itself
What the site said: Wes Virdell’s page gave him 40 out of 100, from seven of his eight records. Two of those seven rows were the same decision: his recorded Yea on Record Vote 3192 of 21 May 2025, on passage of the hemp-THC bill to third reading, counted as -2 from the bill table, and the reason for vote he filed in the House Journal explaining that Yea, counted as -1 beside it. One vote moved the number twice.
What it says now: His score reads 42 out of 100, from six of eight records. The reason for vote is still printed on his page, in full and in his own words; it no longer moves the number, and the row explaining why says so. A statement of vote — a member recording that the vote shown was not the one intended — is separate evidence of a position the roll call misstates, and those are still counted; a reason for vote explains the vote that was cast, which is already counted. Checked by recomputing all 436 candidates’ scores before and after: his is the only number that moved.
How the error was found: An audit of the scoring arithmetic found 36 roll calls where a candidate carried both a vote and a journal statement about the same record number. Thirty-five pointed in opposite directions, which is what a statement of vote means. This was the only one pointing the same way, and that is what singled it out.
Twenty-seven citations to the Texas Legislature bill-author pages were written with a space inside the bill number
What the site said: Twenty-seven source links on ten candidates’ pages pointed at capitol.texas.gov/BillLookup/Authors.aspx with a space inside the bill number — "Bill=SB 90" — which a browser sends as "Bill=SB%2090". The other 390 links to that same page on this site write it "Bill=SB90", and the two are different query values, so a reader clicking one of the twenty-seven had no assurance of reaching the author list the citation was given for.
What it says now: All twenty-seven are written the way capitol.texas.gov writes them itself: "Bill=SB90". Checked against the author pages for all twenty-two bills concerned, which this project already holds in its own cache from an earlier read: on every one of them the page’s own navigation links and form target, generated by the Legislature’s server, spell the bill number without a space. capitol.texas.gov asks automated tools not to read /BillLookup/, so nothing was fetched again to establish this.
How the error was found: A sweep comparing every citation to this one endpoint found twenty-seven written one way and 390 the other, and the inconsistency rather than a fetch is what identified them.
The score explainer promised a reason for every unscored act, and 882 of those reasons were never printed
What the site said: This site shows some acts as evidence and deliberately keeps them out of a candidate’s score — a vote on concurring in the Senate’s changes, a questionnaire answer that points both ways, a bill on the unscored list. 882 such acts were published across 162 candidates when this was measured, and 883 across 163 as it is written, and the data schema refuses any stance that arrives without a written reason, so every one of the 882 had one. None of them reached a page. The reason was validated and then dropped: the only place the renderer ever printed it was the audit list inside the "How this score is worked out" disclosure, and an unscored act never produces a row in that list. The explainer directly above that list told the reader, on all 249 scored candidate pages, that "acts shown with no number were examined and deliberately left out of the arithmetic, with a reason given". The reason was not given. On /candidate/nate-sheets/ a score of 0 out of 100 was published while the reason his 2026 Dallas Morning News questionnaire answer was excluded — that the same answer both backed expanding the Compassionate Use Program and backed enforcing the federal closure of the hemp-derived THC loophole, so no single direction can be read from it — existed in the data and appeared nowhere a voter could read it. The reasons on acts that DO carry a direction reached the page only as the label of a row in that same disclosure, detached from the act they explain and behind a click.
What it says now: Every published reason for leaving an act out of the arithmetic is printed on the card of the act it explains, on the candidate page and on the race page, in the open page rather than in a disclosure or a tooltip. All 883 appear; none is missing or empty. Because a reason for exclusion is this site’s own judgment and not a thing the candidate said or did, each one is set apart from the record it sits under and carries its own byline, "WeedVoter’s note — shown here, but deliberately left out of the score", so it cannot be read as more of the record of a named living person. The explainer’s sentence is now true. tests/stance_why.mjs fails the build if any written reason is absent from the visible markup of its candidate page, strips every title attribute before measuring so a tooltip cannot satisfy it, names the candidate whose reason is missing, and pins the Nate Sheets case by name.
How the error was found: A review asked whether the site keeps the promises its own explainers make. The claim was checked against the 436 built pages rather than against the code, by extracting the visible text of each page — title attributes removed, because a tooltip is not publication — and searching it for the exact reason stored in the data. The answer was 0 of 882. The same measurement found a narrower defect that is not fixed here and is recorded as outstanding: an act that does carry a direction loses its written reason whenever the act is tied to a bill, because the score row then shows the bill caption in place of the reason. One act is in that state today, Colin Allred’s 2023 cosponsorship of H.R. 2891, and the reason it scores +1 rather than +2 is on no page.
Candidate pages promised records "below" that were not on the page
What the site said: Every candidate page opens with the blurb for the race it belongs to. 141 of the 215 race blurbs point the reader at records "below" — "all eight record votes are listed below in date order" and similar — because on the race page the records are below. On 87 of the 436 candidate-page sections that paragraph sat above a card holding no records at all, so the page told a reader that a named living person had a documented cannabis record and then showed none. On /candidate/alexandria-nicole-butler/ the first paragraph promised eight record votes and a statement of vote "below" and the card under it read "No record compiled yet". A scope note had been added to 59 of the 87, but it arrived in the third paragraph and left the word "below" standing in the first. On /candidate/ron-c-angeletti/ there was no note at all: the Senate District 4 blurb describes the person who won the May 2026 special election and "currently sits in the Texas Senate", a seat-holder sentence in none of the words our guard looked for, and he is not himself on the November ballot, so the name check did not catch it either.
What it says now: Where a candidate page carries no record list, the blurb’s "below" is re-aimed at the race page, where the records it promises actually are: "listed below in date order" now reads "listed on the race page in date order" on those pages, and nowhere else. Nothing else in the paragraph is changed — no fact, no name, no date — because "below" is the only word in it whose meaning depends on which page it is printed on. The paragraph that says so in words is printed on those pages too, naming the candidate whose page it is and linking the race page, and the guard that decides when to print it now recognises "currently sits" as a sentence about the holder of the seat. No rendered section promises records below a card that has none; tests/promise_below.mjs fails the build if one does.
How the error was found: A review asked whether the pages promise a record they do not contain. The counts were then measured against the 436 built pages rather than the generator, which is where the defect existed: the code that printed the blurb was correct code doing the wrong thing.
A double negative inverted a finding on a judicial candidate’s page
What the site said: The search log for Audra Riley, a candidate for the Court of Criminal Appeals, said of two sites belonging to a judge of that name: "Neither contains no occurrence, on a word boundary, of cannabis, marijuana, marihuana, hemp, THC, CBD, Compassionate Use, delta-8, delta-9, dispensary, decriminalize, cannabinoid or psilocybin." Read as written, that says both sites DO mention those words — the opposite of what was found. It appeared on her candidate page and on the Court of Criminal Appeals, Place 4 race page.
What it says now: It reads "Neither contains any occurrence". The finding is unchanged: both sites were read in full and neither mentions any of those terms, and both are 2024 sites for a different office, so neither is treated here as her campaign site for this race.
How the error was found: A review of what the pages assert about their own subject.
A House District 93 candidate had stopped campaigning and our pages did not say so
What the site said: The House District 93 pages described a contest between two candidates and said the same two candidates were standing in both the general election and the special election for the unexpired term. They did not mention that the Democratic nominee, Ericka Lomick, announced in September 2026 that she had suspended her campaign. A voter reading us thirteen days before early voting would have had no way to know.
What it says now: Both pages say so now. They quote the Tarrant County Democratic Party’s statement of 22 September 2026 — the party that nominated her — and they explain why her name is printed on both ballots anyway: the announcement came after Friday 21 August 2026, which the Secretary of State’s election law calendar for this election gives as the last day a withdrawal could have caused her name to be omitted. Tarrant County Elections Administration has said she remains on the ballot. Her answers from February 2026 are unchanged and still shown, and the page now dates them so a reader can see they were given before the announcement. ⚠️We have not published, and do not publish, the reasons others have given for the suspension. Material of that kind about a named living person needs a court or state record, not reporting of what her opponents say.
How the error was found: A sweep for candidates whose public campaign activity had stopped since their page was written. Nothing in this pipeline could have caught it: every job here reads filings, roll calls and bill histories, and a suspended campaign leaves a filed candidacy looking exactly like an active one. The signal only exists in party and news statements, which nothing was reading.
We published a zero out of a hundred from a source we were not allowed to read
What the site said: Nate Sheets, a candidate for Agriculture Commissioner, was published at 0 out of 100 — the lowest score this site gives. The whole number rested on one entry describing what he said in a recorded video interview, with no quotation of him in it. The publisher of that interview asks automated tools by name not to read its site, so we had no permitted way to check the description. The one other source cited on that entry, which does permit us, neither quotes him nor says the same thing: it reports his criticism of the hemp industry as being that it has focused on smokable products, and his interest in uses such as paper and clothes and in crop rotation.
What it says now: The description of the interview is removed, because we should not have published an account of a source we were not allowed to read — the same rule we applied this week when we withdrew quotations taken from a state website that refuses us. The entry now reports only what the permitted source says, says plainly that those are the reporter’s words rather than his, and records that the interview is cited but not relied on. With that gone the entry carries no direction, so he no longer has a score at all. ⚠️This is not a judgement that he is more supportive than the number said. Nobody here has decided that. We removed a number we could not stand behind, and a candidate with nothing we can verify gets no score rather than the worst one.
The Austin appeal we said we could not trace has been found
What the site said: Our Austin entry said the state had filed a separate appeal against the temporary injunction and that “we have not been able to confirm how that appeal ended, because the only court docket that would show it refuses automated access”.
What it says now: It has been found, on a court-document host that does permit us. The appeal was No. 15-26-00127-CV, taken from a temporary injunction signed 2026-05-05. The state moved to dismiss its own appeal on 2026-08-10 and the court dismissed it on 2026-08-18 without reaching the merits, so the appeal decided nothing about the ordinance. The entry now says that, and also gives the date the injunction was signed. ⚠️We have not concluded anything further from the dismissal: it does not show that the trial court ever entered the wider injunction the appeals court directed, and no order we can read says it did. A sentence that had been printed twice in the same entry has also been removed.
Seventy-four cards said less than what we had actually checked
What the site said: Of the 175 candidates with no cannabis record, 74 had no reading of their own campaign site recorded on their card at all. Some of those cards rested on a stored sweep that had logged a content delivery network’s refusal — an image or font host, nothing to do with the candidate — as though it were a fact about the person. One card said a candidate’s issues page had been “read in full” and named its three headings without recording any address for it.
What it says now: All 74 are rewritten, and they now say which of three different things happened, because those are not the same claim. For 25, we found the candidate’s own site, confirmed it is theirs by the name and office printed on it, read every page we could reach, and name the page count and the terms searched. For 8 we found their site and could not read it — seven because the host refuses automated reading and one because the page carries no readable text — and those cards say plainly that an absence of evidence we were not allowed to gather is not a finding. For 41 we searched and found no site of their own at all. Seven cards also record a site we found and rejected because it belongs to somebody else of the same name, including one in California whose pages discuss cannabis and which is not attributed here in any form. The right-of-reply mailing list has been rebuilt from the corrected cards.
Our methodology page described a different method from the one the code runs
What the site said: The methodology page said “only two kinds of act move the number” — a recorded vote and authoring or signing on to a bill — and that statements, questionnaire answers, interviews and campaign platforms “do not move the number”. The same sentence was printed inside the score box on every scored candidate page. The page also promised the arithmetic could be redone by hand, and then printed only three of the six weights.
What it says now: That was false, and it had been false the whole time. Statements and questionnaire answers have always counted, at a lower weight than a recorded vote; an executive action counts the same as a vote and an organisational endorsement counts least. Of the 250 candidates carrying a score, 152 have numbers that include such acts, and for 77 of them — usually challengers who have never held office and so have never cast a vote — those acts are the entire score, including all three of the scores at zero and 48 of the scores at 100. Nobody could have reproduced those numbers from the weights we printed. The page now lists all six weights, says which of them rest on our reading of prose rather than on a roll call, and says plainly that the earlier version was wrong. No published number changed: the description was wrong, not the arithmetic. Whether the scoring itself should change is a separate question we have not yet decided. ⚠️A first attempt at this correction, earlier the same day, was itself wrong. It fixed one section and left five sentences elsewhere on the page saying the opposite, so the page contradicted itself; it mis-stated the number of kinds of act as six when there are seven; and the replacement sentence it put on every scored candidate page contradicted the sentence immediately before it and ended without a full stop. All of that is now fixed. The sentence that mattered most was the one defining a zero as “the score of someone who voted consistently to restrict access”: three of the people at zero have never cast a vote on this subject, because they have never held office. That page now says what a zero does and does not mean.
Nine short descriptions above the records counted those records wrongly
What the site said: Each race page opens with a short description of what the candidates’ records contain. Nine of them stated counts the records underneath flatly contradicted — one said a member had cast “four record votes” where the page lists 28, another “nine votes between 2013 and 2025” where the page lists 26, another “all three entries” where the page lists 17. The worst said the journal recorded “no statement of vote on any of them” while the statement of vote was printed on the same page, directly below. One more denied that a member had authored or coauthored anything, above two coauthorships.
What it says now: All nine are recounted from the records themselves and rewritten. Where the original made a narrower point that was true — that a member was absent for three of the four floor votes on the hemp ban and the medical expansion, and voted Nay on the fourth — the point is kept and scoped to those votes rather than flattened into a total. The test that is supposed to catch this had been recording these disagreements as known and passing anyway; it now fails on them, the list of accepted exceptions must be empty, and a sweep of all 215 descriptions found the two further errors included above. We are sorry: these sat above the evidence that disproved them.
We said the San Marcos ordinance was blocked. We had not checked that it is.
What the site said: Our city-by-city table labelled San Marcos “Blocked in court” and said the ordinance “is blocked while the case is tried rather than struck down”.
What it says now: The Fifteenth Court of Appeals ordered the trial court to enter a temporary injunction, but its own opinion records that at that point “there is no temporary injunction in place”, and we have found no later order showing one was entered. An order telling a court to do something is not proof it was done. The row is relabelled “In litigation”, which is what we can show: the case is live, no trial has been held, and no final judgment exists. Austin keeps the stronger label because a 2026 mandamus proceeding there does record that the trial court entered an injunction. The same overstatement appeared in two letters we had drafted to San Marcos newspapers, and both have been rewritten.
We listed a Maine cannabis measure as having failed, when it is still being signed
What the site said: Our list of measures that did not qualify for the 2026 ballot included a Maine measure, described as a repeal of adult-use legalization, with the note “Did not meet the 2026 deadline. Organizers could pursue 2027.” Our source for it was a national advocacy group’s round-up.
What it says now: The Maine Secretary of State lists it under “Citizen Initiative Petitions Approved for Circulation” with a filing deadline of June 8, 2027, and its petition organiser registrations were updated on 2026-09-28. So it is still live and being circulated for signatures, and because its deadline falls after this election it was never capable of reaching the 2026 ballot — it did not fail, and there was no 2026 deadline for it to miss. Its own title is “An Act to Amend” rather than a repeal. The entry now says all of that and cites the Secretary of State’s page instead of the round-up. This is a reminder to us that a national tracker is a lead and not an authority, which is the rule this guide is supposed to follow.
We quoted a state official from a website we are not allowed to read
What the site said: Our Massachusetts Question 8 page carried several sentences in quotation marks attributed to “the Secretary of the Commonwealth’s own words”, including the official statements of what a yes and a no vote would do. Its first listed source was an archived copy of the Secretary of the Commonwealth’s website.
What it says now: That website refuses automated reading, so our own rule forbids us from reading it — and reading it through an archived mirror is exactly what that rule forbids. We had not read the page those quotations came from, so we should never have printed them as quotations. They are removed, along with the archived link. The page now draws on the Attorney General’s official summary of the measure, printed in the Legislature’s own copy of the petition, which we can and did read, and it says plainly which statements are the state’s and which are ours. Nothing we said about what the measure does has changed.
Two contests on the November ballot were missing from this guide
What the site said: The guide listed 215 races. It did not include the special elections for the unexpired terms in Texas Senate District 22 and Texas House District 93.
What it says now: Both are on the November 3 ballot, ordered by the Governor after the sitting members resigned, and both are now listed. Voters in those two districts will see the same candidates twice — once for the few remaining weeks of the old term and once for the full term beginning in January — and the pages now say so. We found these by checking the Secretary of State’s own list of special elections against our ballot; they are not in the party-nominee certification we had reconciled against, which is why they were missed.
We placed a candidate’s remarks at the wrong event, on the wrong date
What the site said: An entry on Vikki Goodwin’s page said she made remarks about taxing THC like alcohol “at a Houston rally, as reported that day”, dated 2026-09-04.
What it says now: The quotation is accurate and hers, but the setting and the date were not. The outlet that carried the words published them on 2026-08-25 and places the remarks a week earlier still; her Houston rally happened afterwards. The entry is re-dated to the day it was reported, the claim about the rally is withdrawn, and the entry now says the date is a reporting date rather than the day she spoke.
Two reasons for leaving a bill out of the arithmetic described the bill wrongly
What the site said: On the page listing which bills we score, 88R HB 4918 was described as “arguably expansive and arguably a restriction on an existing market”. Separately, 89R HB 513 was described in a way that read as though cannabis were incidental to it.
What it says now: Reading the engrossed text of HB 4918, it restricts nothing: its three operative provisions strike the word “not” from Agriculture Code 122.301(b) and repeal Health and Safety Code 443.204(4) together with the 443.001(11) definition that provision relied on. It is a repeal of a ban on one product form, and the reason now says that. For HB 513, marihuana is named in the statute the bill amends, so the reason now says the judgement rests on the bill’s trigger being a death or a serious bodily injury rather than on cannabis being absent. Both bills are still outside the arithmetic and no published number changed, so every vote on them appears on the candidates’ pages exactly as it did before. Whether HB 4918 should now be scored is a question we have not settled, and the reason text says so rather than implying it is closed.
We predicted a pattern of Oregon marijuana tax measures. We went looking, and it is not there.
What the site said: Our entry on Warrenton, Oregon said its marijuana tax measure was “the clearest example of a pattern worth watching: Oregon’s 3% cap is sending cities that adopted higher rates before it back to their voters.” That was a prediction about measures we had not found, printed on a page whose whole claim is that it only states what it can show a reader.
What it says now: The prediction was tested and did not hold. A statewide round-up of local tax measures dated 2026-08-26 lists five on this year’s Oregon ballot and Warrenton’s is the only cannabis one, and twelve county measure lists read directly carry no second marijuana question. The entry now says that: it is the only local marijuana tax measure in Oregon this cycle. What a yes or a no vote does is unchanged — a yes still lowers the city’s rate from 10% to the 3% state law now allows.
We told 36 candidates we had read their Ballotpedia page. We had read the wrong page.
What the site said: The “what we checked” card for 36 candidates said their Ballotpedia profile had been read in full and contained no mention of cannabis, and some gave a character count as proof. For every one of those 36 the page we had actually fetched was either a page that does not exist or a disambiguation page listing several different people of the same name. The character counts measured those wrong pages. Nine of the cards also said the candidate had completed no Candidate Connection survey when they had.
What it says now: Every one of the 36 correct profiles has been found and read, and all 36 are still clear of any mention of cannabis, hemp or THC — so the conclusion survives; what was wrong was the evidence for it. Each card now cites the correct address, says plainly that an earlier check cited the wrong page, and says where a survey exists. Two further candidates, Howard Olsen and Denise Wooten, appear to have no Ballotpedia article at all; their cards will say we could not check rather than that we checked and found nothing.
The most consequential federal cannabis vote was on every page and in no score
What the site said: In November 2025 the House voted to enact the law that narrows the federal definition of hemp. We published that vote on the pages of 24 Texas candidates, with a direction attached to each one — and the scoring code silently discarded all 24, because a recorded vote takes its direction from this site’s table of bill rulings rather than from the record itself, and that bill was missing from the table. So the vote was displayed on 24 pages and counted on none of them.
What it says now: The bill is in the table, recorded as a government funding act whose cannabis section no member could vote on separately — the same treatment as the three other omnibus funding votes on this site. The 24 records now say in words that they are shown and not scored, and why. The scoring code also refuses to read a direction from a bill marked as excluded, which previously would have turned a candidate’s score into a non-number.
Houston Proposition A now quotes the certified ballot, and our signature figure was loose
What the site said: Our entry for the only Texas cannabis measure on this ballot took the proposition letter from news coverage dated 2026-09-29 rather than from a certified ballot, because the City of Houston, the City Secretary, the city agenda system and the Harris County sample-ballot host all refuse automated reading. It described only two of the amendment’s eleven sections, omitted the caption’s own phrase "with related reporting requirements", and said the campaign gathered "more than 40,000 signatures" without saying that figure was a count before verification.
What it says now: The entry quotes the certified ballot text in English and Spanish, as printed by two independent county election authorities — Montgomery County’s master sample ballot and 41 of Fort Bend County’s precinct ballots, which carry the identical caption. All eleven sections are described, including the five-year enforcement audit and the annual reports the caption refers to. The signature figures are now the campaign’s own: 22,800 valid against 20,000 required, from about 40,000 collected before verification. A reported City Secretary sample finding 57 per cent valid is not stated, because no official document we could reach supports it. Press summaries describe the measure as barring the smell of marijuana alone as grounds for a search; those words appear nowhere in the charter text or the certified caption, so this guide does not state it.
We printed a candidate’s initials as a word
What the site said: The page for the Texas House District 30 candidate read "Aj Louderback". The name was ingested in capitals from a legislative journal and a title-casing step turned the initials into a word.
What it says now: It reads "AJ Louderback", which is how the Secretary of State’s certified candidate list prints it. We checked the other 435 names on the ballot against that list for the same fault and found none.
We said two courts had struck down the Austin and San Marcos ordinances. They had not.
What the site said: The Austin row of our city table said the Fifteenth Court of Appeals ruled "the ordinance conflicts with state law and cannot be enforced", and Ken Paxton’s page carried one record covering "the Austin and San Marcos cases" to the same effect. Both overstated what the court did, and the single record also merged two separate decisions — San Marcos on 2025-04-17 and Austin on 2025-04-24, by different authoring justices — into one.
What it says now: Neither decision was a final ruling that the ordinances are void. In each the court reversed the dismissal and the refusal to block the ordinance and sent the case back with an instruction to enter, in the Austin court’s own words, "a temporary injunction prohibiting the enforcement of the Ordinance pending a final trial on the merits". So they are blocked while the cases are tried. The two decisions are now described separately with their cause numbers and reporter citations, the Supreme Court of Texas outcomes are given for both, and the 2026-07-30 mandamus holding that the Austin injunction must also cover the city’s THC-testing spending has been added. We read the opinions on a source that permits automated reading, because the Texas judiciary’s own opinion search asks tools not to.
A sitting member of Congress was shown as a candidate, and a candidate as the sitting member
What the site said: Christian Dashaun Menefee was listed as a nominee in United States House District 18, although he won the special election there on 2026-01-31 and was sworn in on 2026-02-02. In District 37, Greg Casar was labelled "Incumbent", although the Clerk of the House lists Lloyd Doggett for that district and Mr Casar for District 35.
What it says now: Mr Menefee is shown as the incumbent. Mr Casar is shown as a nominee, and the District 37 page now explains that he is a sitting member of Congress for a differently numbered district under the map now in force and is standing here under the map being voted on in November. That is the same mid-decade redistricting this site was built to explain, and we had published it backwards on our own page. Every one of the 38 congressional races was re-checked against the Clerk’s roster.
One questionnaire answer was counted twice, and one receipt for a no-record finding was a dead link
What the site said: Angie Carraway’s answer to The Dallas Morning News’s 2026 voter-guide question appeared twice on her page — once citing the newspaper’s live page and once an Internet Archive copy of the same data file — so a single answer was presented as two records. Separately, the Community Impact candidate questionnaire cited on Helen Zhou’s page, which is the evidence for our finding that no cannabis record was found for her, returned a 404.
What it says now: The answer is one record, citing the live page first and the archive copy as a backup. The Community Impact link is corrected — the address had "qa-" where the paper uses "q-a-" — and the restored page confirms what the card says: published 2026-09-04, and no mention of cannabis, hemp or THC anywhere in it.
We labelled a candidate the incumbent of a seat he does not hold
What the site said: The page for Colin Allred, a candidate for the United States House in District 33, carried the label "Incumbent". He left the House in January 2025. The Clerk of the House lists a different member for that district, and he appears in none of the roll calls of the current Congress. Two separate checks of the federal record flagged it independently before we caught it.
What it says now: He is listed as a nominee. His earlier votes and cosponsorships from the years he did serve are unchanged and still shown, because they are his record; what was wrong was the claim that he holds the seat now.
The election countdown showed Texans one day fewer than they had
What the site said: The counter on the front page worked out today’s date in UTC rather than in Texas time. After 7 p.m. Central the UTC date has already rolled over to tomorrow, so from early evening onwards a reader in Texas was told there was one day less until election day than there actually was. On a countdown to a voting deadline, in the weeks when people are deciding whether they still have time, that is the worst direction to be wrong in.
What it says now: Both the countdown and the new voting-deadlines pointer work the date out in America/Chicago. A test now fails the build if either one goes back to reading the UTC calendar day.
We said we had read a candidate’s campaign site. It was not his site.
What the site said: The card for Ronald Dwayne Whitfield, a candidate for the United States House in District 18, said "THE CANDIDATE’S OWN CAMPAIGN SITE WAS READ ON 2026-09-20 at texpolls.com — found by search and confirmed to name this candidate before it was attributed to them", and then reported that nine pages and the site’s sitemap had been checked for cannabis content. texpolls.com is not his site. It is an independent Texas elections tracker, and the only place it mentions him is a row in a table of race matchups. Our own rule is to confirm that a site names the candidate before attributing it to them; the check passed on a page that merely listed him, which is the exact mistake the rule exists to prevent.
What it says now: The claim and everything that followed from it are withdrawn. The card says that no campaign site has been found for this candidate, and says plainly that we got this wrong before. The rest of what was checked for him — his Ballotpedia profile, the Texas Legislature roll, the one outlet that put the question to congressional candidates — stands, and none of it mentions cannabis.
We called 37 congressional districts "a Texas legislative district", and told readers federal votes were Texas roll calls
What the site said: Every one of our 38 US House race pages but one opened "A Texas legislative district." Twenty-five of them went on to say that the records listed below were the sitting member’s "recorded votes on the 2025 hemp-THC ban and the Compassionate Use Program expansion … with their record-vote numbers, as the official journals print them." For 22 of those races not a single record is a Texas roll call. They are votes and cosponsorships in the United States Congress. We named a source the records did not have, on 119 pages, about members of Congress who never cast a vote on Senate Bill 3.
What it says now: The congressional pages say "A United States House district" and describe what the records actually are. Where a race mixes federal records with Texas ones, because a sitting state legislator is running for Congress, the page says so instead of denying either. The cause was a script written for the Texas House and Senate and then reused for the congressional races without changing its wording; it now branches on the office rather than assuming the chamber, so a rebuild cannot put the error back.
Seven races said no cannabis record existed, directly above records that did
What the site said: Seven race pages carried the sentence "No candidate on this ballot holds Texas legislative office, so none appears in any 2025 cannabis or hemp roll call" while candidates on the same page had cited records shown underneath it. The sentence was literally true — those records are questionnaire answers and statements, not roll calls — but a reader would fairly take it to mean nothing had been found.
What it says now: Those pages say that no candidate appears in a 2025 roll call and then say what is below: what each candidate has said or answered, in date order, each with its own source. One of the seven was Texas House District 15, which only became misleading when we published a statement by a candidate there earlier the same day — adding a record made an existing sentence wrong, which is a failure mode we had not been checking for and now assert against.
Our page on the national ballot measures rested on trackers and Wikipedia
What the site said: Of the ten cannabis measures this site tracked outside Texas, eight cited no official record at all. The Massachusetts repeal — the biggest cannabis question on any ballot in the country this year — was sourced to two advocacy news sites and Wikipedia. This site tells readers that a claim without a primary source is worth less than one with it, and then did not meet that standard on its own national page.
What it says now: Every measure now cites an official record. For Massachusetts that is the initiative petition as the Secretary of the Commonwealth transmitted it to the Legislature, which carries the full text, the Attorney General’s certification and her official summary; for Idaho it is the Secretary of State’s own voter pamphlet. Both states’ election websites refuse automated reading, which is why this took so long: the Massachusetts document was reached through the Legislature’s copy instead.
We understated what a yes vote does in Massachusetts, and overstated what one does in Idaho
What the site said: The Massachusetts entry said the measure ends "licensed retail sales and commercial cultivation" and that "limited personal possession remains allowed". It did not say that growing at home for non-medical use would also end, and it did not say the repeal would not take effect until 2028. The Idaho entry said the amendment would stop "citizen initiatives and local governments" from legalising marijuana.
What it says now: The Massachusetts entry says the repeal reaches non-medical home growing, quoting the petition’s own statement that it repeals the chapters governing the "possession, use, distribution, cultivation, and taxation of marijuana not medically prescribed", and gives the effective date of 2028-01-01. The Idaho entry no longer mentions local governments: that phrase appears nowhere in the ballot question, the amendment text, or either of the state’s official statements about what a yes or no vote does. It was our inference, printed as though it were the measure’s effect. Both entries now quote the state’s own wording first.
Three wrong facts about measures that did not make the ballot
What the site said: We said a Washington initiative to cut the cannabis excise tax "fell short on valid signatures"; that Florida’s Division of Elections "verified 783,000 of the 880,000 required" and that "a court denied the campaign’s challenge"; and that Missouri had certified seven measures for this election.
What it says now: The Washington initiative never submitted any signatures — the Secretary of State’s list of every petition handed in this cycle does not include it, and being rejected and never filing are different things to say about a campaign. Florida’s requirement is 880,062 exactly, and the only court action in the state’s own record is a dismissal on 2026-02-04 of the advisory review of the petition; the 783,000 figure appears in no state record we could reach, so it is gone. Missouri has certified five measures, counted on the Secretary of State’s page; none concerns cannabis, which was the point the entry was making and which still holds.
The city table said Elgin’s marijuana ordinance was settled. Its council had repealed it.
What the site said: The row for Elgin said the Attorney General’s suit over the ordinance Elgin voters approved in 2022 "was resolved by consent decree in June 2024", and stopped there. It had been overtaken. An Elgin voter reading that page would have thought the ordinance they voted for still stood in some form.
What it says now: The row says the ordinance was repealed. The City Council repealed it on 2026-07-07 under the article of the city charter that allows a voter-approved initiative ordinance to be repealed after two years, and the city attorney told the council on the record that the repeal would make the Attorney General’s suit moot. This was checked in Elgin’s own minutes rather than in news coverage. The minutes record the motion carrying with "All voted in favor" and name no individual votes, so the row does not either.
A senator’s statement was filed under the date it was reported, not the date he made it
What the site said: The most recent act on the site — Senator Charles Perry’s remarks on hemp-derived THC at a Texas Senate interim hearing — was dated 2026-07-14. That is the day Spectrum News published its report of the hearing. The hearing was a week earlier. This site’s rule is that an act carries the date of the act, and dating one to the day a newspaper noticed it puts a candidate’s record in the wrong week.
What it says now: It is dated 2026-07-07, the date on the Senate’s own minutes for the Health and Human Services Committee: "Tuesday, July 7, 2026, 10:00 AM, Capitol Extension, Room E1.012." The official notice and minutes are cited alongside the report. The quotes are still attributed to the report that carried them, because the minutes record the hearing and not the remarks.
Thirty more bills carried a candidate’s name without crediting them
What the site said: Texas Legislature Online prints the author of a bill as a bare surname. Because this site has attached one person’s record to another of the same name before, every such name was held back until it could be settled, and seventy-six were still waiting. Among them were four hemp and cannabis bills a sitting senator filed himself in the 2025 called sessions, and a possession bill a House member filed in 2012.
What it says now: Thirty are published and forty-six are recorded as belonging to someone else. What settled them is a convention of the Legislature’s own: its roster for each Legislature prints members exactly as its bill pages do, and in all nine chamber rosters checked every printed name is unique — so a bare surname is the state saying one member of that chamber bore it that session. Each name was matched against the roster of the chamber the bill was filed in, and every primary authorship was confirmed a second time against the bill’s own history page, which names its author. The forty-six rejected include a senator and a representative who share the surname King, another pair who share Cook, and a Paxton who is not the one on this year’s ballot.
Legislators who wrote a bill were credited only with voting on it
What the site said: For several cannabis bills the site published the floor votes and not the names printed on the bill itself. Members who joint-authored or co-authored the 2021 and 2023 Compassionate-Use expansions, the overdose defence, the 2019 hemp bill and the called-session hemp and cannabis bills appeared only as having voted. Putting your name on a bill is a stronger statement than voting for it, and this site weights it the same, so those records understated what the Legislature’s own records show.
What it says now: Sixty-one filings, joint authorships and co-authorships are published, each with the date the Legislature records against that member’s name. Every one was checked against the chamber the bill was filed in and who actually sat that session: a Senate bill’s author is a senator, and several names that match a candidate’s surname belong to different people who served years earlier.
Bills filed in the other chamber were missing from candidates’ records
What the site said: A Texas bill is usually filed twice, once in each chamber, and Texas Legislature Online names the other chamber’s version on every bill page. The site had recorded the House version of many cannabis bills and not the Senate one, so senators who filed the companion themselves had no record of having done so. Nathan Johnson had filed five that were missing; Charles Perry four; Sarah Eckhardt four.
What it says now: Twenty-one companion bills are now in the scoring table, each recorded with the same direction as the bill the state itself declares it identical or similar to, and twenty-three filings and co-authorships are published with the dates the Legislature records for them. Three names were checked and rejected as different people: the 2015 coauthors named Allen, Isaac and Peña are not the candidates of those surnames on this year’s ballot.
Both House votes on the 2021 THC penalty bill were missing
What the site said: House Bill 2593 of 2021 became law. It moved possession of tetrahydrocannabinols into a new penalty group whose lowest tiers are misdemeanors, out of one where any amount was a felony. The site knew the bill — two candidates were shown as having filed or signed on to it — but published neither of its recorded votes, so seventy-two candidates who voted on it twice had no record of either vote.
What it says now: Both are published: final House passage on 2021-04-28, which carried 108 Yeas to 33 Nays, and adoption of the conference committee report on 2021-05-30, which carried 95 to 44. Statements of vote filed on each day are published with them. Every sweep until now asked whether a bill was known; a new check asks, of every bill the site already cites, whether a recorded vote exists on a day we publish nothing. The same check then found four more missing roll calls, all now published: the Senate’s passage of that bill on 2021-05-21 by 24 Yeas to 7 Nays, the Senate’s vote on Senate Bill 6 of the second called session on 2025-08-18 by 22 to 8, and both House votes on House Bill 4879 of 2025.
A Senate vote on barring local cannabis ballot measures was missing
What the site said: The site carried no record of the Texas Senate’s votes on Senate Bill 1870 of 2025, which would bar a city or county from putting a measure like the one on Houston’s November ballot to its voters at all. Only a single House cosponsorship of that bill was published, so fourteen candidates who voted on it — nine for and five against — had no record of it.
What it says now: Both recorded votes are published: passage to engrossment on 2025-04-22 and third-reading passage on 2025-04-23, each 23 Yeas to 8 Nays with the names printed in the Senate Journal. Each senator was matched to their district in the Senate roster rather than by surname, and the Yeas and Nays together were checked against the roll of senators present that day before any vote was attributed.
Bills filed by candidates were shown as evidence but counted for nothing
What the site said: Eighty-four published acts — nearly all of them bills a candidate filed, authored or signed on to — cited bills on which this site had never recorded a judgment. An act whose bill has no entry in the scoring table counts for nothing, so those acts were printed under a candidate’s grade while contributing none of it. Because authorship is weighted the same as a floor vote, this fell hardest on the members who write cannabis bills: a legislator who filed a legalisation bill every session and voted once had only the vote counted.
What it says now: Every one of those bills now carries a recorded judgment — eighty-four scored and twenty-four deliberately not scored, each with its reason published. Where a bill’s caption stated only the subject, such as “relating to the criminal penalties for possession of marihuana”, which fits a bill that raises penalties as well as one that lowers them, the introduced bill text itself was read and the ruling cites it. A test now fails the build if any published act cites a bill with no ruling.
Committee votes were shown as evidence but left out of the score
What the site said: Thirty-four votes in Texas House committees to report a cannabis bill favorably were published on candidates’ pages, under the evidence that explains their grade, and were not counted in the grade itself. The scoring code recognised a committee report only when fewer than sixty characters separated the word “report” from “favorably”, and the sentence these acts carry puts the bill’s full description in between. Twenty-four candidates across seven bills had a score computed from fewer records than their own page listed.
What it says now: All thirty-four are counted. The scoring page and every affected candidate’s number were recomputed, and a test now checks each wording the pipeline produces rather than only the two that had been written by hand.
Four cannabis amendment votes were missing from twenty members’ records
What the site said: The site published the House’s cannabis amendment votes for only two candidates. Because the Clerk of the House titles an amendment roll call by its sponsor and never by its subject — “Mast of Florida Part B Amendment No. 28” — there was no way to search for the rest, and they had been found by hand or not at all.
What it says now: Every amendment offered to a bill that had an amendment roll call between 2021 and 2026 was read against the Library of Congress’s own description of it — 5,021 amendments — and the eleven that concern cannabis were examined by hand. The six put to a recorded vote are now published where they belong: the 2021 amendment on federal marijuana research funding, the 2022 amendments on security clearances and on two studies, and the 2024 amendment on Veterans Affairs forms. The two study amendments are printed but not scored, because a study takes no side. An amendment offered to the MORE Act is now scored on its own terms rather than the Act’s.
Eighteen members of Congress were missing two 2022 cannabis votes
What the site said: The records of eighteen sitting members of the Texas delegation ended at the MORE Act vote of 2022-04-01, as though the House had taken no cannabis vote since. It had taken two more that year, and most of these members voted for both, so the omission made their records read as less supportive than the roll calls show.
What it says now: House Roll Call 108 of 2022-04-04, passage of H.R. 5657, the Medical Marijuana Research Act, and House Roll Call 392 of 2022-07-26, passage of H.R. 8454, the Medical Marijuana and Cannabidiol Research Expansion Act, are now published for every one of them. Each vote is taken from the Clerk of the House’s own roll-call XML and matched to the member by bioguide id, not by surname; the counted votes were checked against the Clerk’s printed tally before any were attributed.
A 2015 Senate vote was described as the wrong vote, with the wrong tally
What the site said: Don Huffines’s record said the Senate passed the 2015 Compassionate-Use bill “on second reading” with him named in the printed Yeas roll, and that it later passed third reading by 31 Yeas to 0 Nays.
What it says now: The Senate Journal for 2015-05-07 records the second reading as a viva voce vote with no roll, naming only five Nays, and the third reading as 26 Yeas to 5 Nays with both lists printed. His record now says that, and the same vote has been added for the seven other candidates named in that roll.
The Texas city-by-city law table was out of date and incomplete
What the site said: The table said seven Texas cities had voted on marijuana enforcement since 2022. It listed Dallas as in litigation after a judge let the city enforce Proposition R, and Denton as in litigation with no outcome confirmed. It left out Harker Heights, Bastrop, Lockhart and San Antonio. The Houston measure page said appeals courts had struck down the ordinances of all five cities sued in 2024.
What it says now: Dallas agreed in June 2025 to stop enforcing Proposition R while its case continues, and Denton’s council repealed its ordinance on 2025-05-20; both rows are corrected. Harker Heights, Bastrop, Lockhart and San Antonio are added, each with sources, and the table now counts ten cities from its own rows. The Houston page now says an appeals court ruled on two of the five.
The Land Commissioner race denied records printed below it
What the site said: The race summary said "No candidate in this race has a documented cannabis act", above Dawn Buckingham’s recorded Senate votes on hemp and the Compassionate-Use Program and Benjamin Flores’s published position on the 2026 hemp restrictions. Buckingham, who holds the office, was also not labelled the incumbent.
What it says now: The summary now describes the records on the page, and Buckingham is labelled the incumbent, confirmed on the General Land Office’s own site. A build test now refuses any race summary that says no candidate has a record while its candidates carry one.
Three Texas Senate races showed the wrong incumbent
What the site said: Taylor Rehmet, who has held Senate District 9 since a 2026 special election runoff, was not labelled the incumbent, and the race said no candidate on the ballot holds Texas legislative office. Trent Ashby (District 3) and David Cook (District 22), both members of the Texas House, carried an "Incumbent" label for Senate seats they do not hold, and their races described their House votes as the sitting member’s.
What it says now: All three corrected against the Texas Senate’s own member pages: Rehmet is labelled the incumbent, District 3 is held by Robert Nichols, who is not on the ballot, and District 22 is currently vacant. Ashby’s and Cook’s votes are now described as House votes. A build test now checks every Texas House and Senate race against the Legislature’s roster.
A write-in candidate for Governor was listed as a printed candidate
What the site said: Jenn Mack Raphoon was listed as an Independent whose name is printed on the Governor ballot, and the race said four names are printed.
What it says now: She filed an independent declaration of intent but did not qualify for the printed ballot. She is a certified write-in candidate, listed by the Secretary of State as Jenn Mack. She is now shown with the race's write-ins, and the race correctly says three names are printed.
32 cards claimed a Ballotpedia page had been read when it did not exist
What the site said: 32 "no record found" cards said the candidate's Ballotpedia profile had been read and contained nothing about cannabis. The addresses checked return "page not found"; our checker mistook that page for an empty profile.
What it says now: All 32 claims retracted. Most of those candidates do have a profile, filed under a title without their middle initial. The line was put back only where the profile was confirmed to name Texas and the specific seat, so it is the same person.
Four challengers were labelled "Incumbent"
What the site said: Four US House challengers carried an "Incumbent" label, in races that already had an incumbent: Danny Minton (District 6), Angela "Heli" Rodriguez Prilliman (District 12), Phil Gray (District 26) and Justin Early (District 31). None has served in Congress.
What it says now: Labels checked against the House Clerk's member roster and corrected. A build test now refuses any race with more than one incumbent.
People-search listings were shown as two candidates' campaign websites
What the site said: A people-search directory listing was named as the campaign website of Anthony Gupta, and another as the campaign website of Cheryl Lynn Clark.
What it says now: Both cards now say no campaign website of their own could be found. The directory listings are deliberately not linked, because they carry private details such as home addresses.
112 cards overstated how much of a campaign website had been read
What the site said: 112 "no record found" cards said the candidate's campaign site contained no mention of cannabis, marijuana, hemp, THC or CBD. Only each site's home page had been read.
What it says now: Every site re-read, including the pages its home page links to. Each card now says how many pages were read and that this covers the pages that can be reached, not necessarily the whole site.
A candidate with a published hemp position was shown as having no record
What the site said: Benjamin Flores's card said no cannabis record had been found.
What it says now: The deeper re-read found a question-and-answer page on his campaign site addressing the 31 July hemp restrictions. His record now quotes it, with the source linked.
A card said a candidate for Governor had not been researched
What the site said: Jenn Mack Raphoon's card said her record had not yet been searched, after it had been.
What it says now: The card now lists what was actually read.
The "no score" explanation denied records printed below it
What the site said: Every unscored candidate got the same explanation: no recorded vote or bill authorship. On some pages it sat directly above that candidate's recorded votes, most visibly on the page of the Speaker of the House, Dustin Burrows.
What it says now: The explanation now depends on whether the candidate has records. A build test checks every candidate page for a card that denies a record the same page prints.
Two race summaries contradicted the records beneath them
What the site said: The US House District 2 summary said no candidate held Texas legislative office or appeared in any 2025 roll call, above Steve Toth's 2025 roll calls. The Texas Senate District 4 summary said neither candidate held legislative office, about a sitting state senator, Brett W. Ligon.
What it says now: Both summaries corrected. A build test now checks every summary's claims against the records and incumbency data it sits above.
93 recorded votes were displayed but left out of scores
What the site said: Votes adopting the conference committee report on House Bill 46, the medical cannabis expansion, were shown on candidate pages but not counted in their scores.
What it says now: They are now counted, and the affected scores were recalculated.
Scores ignored candidates' own public statements
What the site said: The first version of the score counted only votes and bill authorships, so it showed "No score" for Dan Patrick and Greg Abbott beside this site's own quotations of their statements on cannabis.
What it says now: Statements now count, each carrying a direction and a written reason for it that is shown on the page, so a reader can disagree with a specific judgement.
Two candidates' questionnaire answers carried the wrong date
What the site said: Leigh Wambsganss's and Taylor Rehmet's answers to a newspaper questionnaire were dated 2026-09-17, the date the page was republished.
What it says now: Dated 2025-10-13, the date the page first carried them, confirmed from an Internet Archive capture now cited as the first source.
One of two records from the Governor's veto proclamation was dropped
What the site said: The veto proclamation for Senate Bill 3 contains two separate acts, the veto and a call for stricter regulation. Only one was shown.
The Governor race said its candidates had not been researched
What the site said: A note on the race said its candidates were not yet researched, after all of them had been.
What it says now: Note removed. The build now refuses a race note that contradicts the records on the page.
Six senators were shown with named votes the Journal does not record
What the site said: Six records said a senator was "recorded Yea" on the two Senate votes on House Bill 46. The Senate Journal prints those votes as 31 to 0 without listing names.
What it says now: Described as unanimous votes of the Senate, which is what the cited page supports.
18 co-authorships carried the bill's date, not the member's
What the site said: 18 co-authorship records were dated from the bill's history, not the date the member actually signed on. One was off by about three months.
What it says now: Re-dated from the Legislature's own list of authors, which records each member's sign-on date.
Corrections carried on candidate cards
275 corrections on 180
candidate pages. Each of these was written into the research note
behind a candidate’s card, which is where it was published until now — inside the one
paragraph a reader reads about that person. The card still says it was corrected and links here.
These carry no date: the day each fix was made was never recorded, and a guessed date on
a corrections log is worse than an absent one.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said Ballotpedia records that they did not complete its Candidate Connection survey. That was our error: the copy of the profile we read on 2026-09-20 already recorded the completed survey.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said Ballotpedia records that they did not complete its Candidate Connection survey. That was our error: the copy of the profile we read on 2026-09-20 already recorded the completed survey.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said Ballotpedia records that they did not complete its Candidate Connection survey. That was our error: the copy of the profile we read on 2026-09-20 already recorded the completed survey.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said this candidate's own campaign site was found and read on 2026-09-21 at alexandriabutlercampaign.org.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile. It also wrongly said this candidate had completed no survey.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card listed the Blue Voter Guide’s page for Texas House District 61, on bluevoterguide.org, as read in full on 2026-10-06, and said it names this candidate with no endorsements, no stated positions and no mention of cannabis, marijuana, hemp or THC; that claim is withdrawn, because when we re-checked on 2026-10-30 the site’s robots.txt asked automated tools, ours among them by name, not to read it, and the copy of its robots.txt we held when it was read has since been replaced, so we cannot show it permitted us then, and nothing from that page is relied on here.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
April FarrisJustice, Fifteenth Court of Appeals, Place 3
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
Audra RileyJudge, Court of Criminal Appeals, Place 4
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
Brett BusbyJustice, Supreme Court of Texas, Place 8
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
Chari KellyJustice, Supreme Court of Texas, Place 2 (unexpired term)
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said no campaign site of their own could be found. That claim was wrong, and it was wrong because the search behind it looked only at search-engine results and never at the addresses other trackers already cite.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this entry rested that identification on the Legislative Reference Library, which asks automated tools not to read it; that citation is withdrawn and the identification now rests on the journal alone.
An earlier version of this entry rested that identification on the Legislative Reference Library, which asks automated tools not to read it; that citation is withdrawn and the identification now rests on the journal alone.
An earlier version of this entry rested that identification on the Legislative Reference Library, which asks automated tools not to read it; that citation is withdrawn and the identification now rests on the journal alone.
An earlier version of this entry rested that identification on the Legislative Reference Library, which asks automated tools not to read it; that citation is withdrawn and the identification now rests on the journal alone.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said Ballotpedia records that they did not complete its Candidate Connection survey. That was our error: the copy of the profile we read on 2026-09-20 already recorded the completed survey.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said this candidate's own campaign site was read on 2026-09-21 at david4texas.com.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said no campaign site of their own could be found. That claim was wrong: the site exists, and the only reason it is unread is that we are not entitled to read it.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile. It also wrongly said this candidate had completed no survey.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile. It also wrongly said this candidate had completed no survey.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said no campaign site of their own could be found. That claim was wrong, and it was wrong because the search behind it looked only at search-engine results and never at the addresses other trackers already cite.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
Holly TaylorJudge, Court of Criminal Appeals, Place 9
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said Ballotpedia records that they did not complete its Candidate Connection survey. That was our error: the copy of the profile we read on 2026-09-20 already recorded the completed survey.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
James P. SullivanJustice, Supreme Court of Texas, Place 2 (unexpired term)
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile. It also wrongly said this candidate had completed no survey.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said no campaign site of their own could be found. That claim was wrong, and it was wrong because the search behind it looked only at search-engine results and never at the addresses other trackers already cite.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
Kyle HawkinsJustice, Supreme Court of Texas, Place 7
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile. It also wrongly said this candidate had completed no survey.
An earlier version of this card listed the candidate’s own campaign site, lauren4congress.com, as having been checked on 2026-09-20 and 2026-10-17 and found silent on cannabis; when we re-checked on 2026-10-20, that site’s robots.txt asked our crawler by name not to read it, and we cannot show it permitted us when those checks were made, so that claim is withdrawn and the finding is not relied on here.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said no campaign site of their own could be found. That claim was wrong, and it was wrong because the search behind it looked only at search-engine results and never at the addresses other trackers already cite.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said no campaign site of their own could be found. That claim was wrong, and it was wrong because the search behind it looked only at search-engine results and never at the addresses other trackers already cite.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
Marc M. MeyerJustice, Fifteenth Court of Appeals, Place 3
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
The earlier note here said his issues page had been "READ IN FULL" and named its three headings without recording any URL, which is a claim about reading a page that names no page.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said no campaign site of their own could be found. That claim was wrong, and it was wrong because the search behind it looked only at search-engine results and never at the addresses other trackers already cite.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card listed the Blue Voter Guide’s page for Texas House District 103, on bluevoterguide.org, as read in full on 2026-10-06, and said it names this candidate with no endorsements, no stated positions and no mention of cannabis, marijuana, hemp or THC; that claim is withdrawn, because when we re-checked on 2026-10-30 the site’s robots.txt asked automated tools, ours among them by name, not to read it, and the copy of its robots.txt we held when it was read has since been replaced, so we cannot show it permitted us then, and nothing from that page is relied on here.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said their profile carried no Candidate Connection survey, although the same card also said their survey answers were published on that page. That was our error: Ballotpedia states the answers were first submitted on 2026-06-03 and updated on 2026-09-01.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile. It also wrongly said this candidate had completed no survey.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this entry described what he said in that interview. We should not have done that, because we were not allowed to read it and could not check it, and nothing from it is relied on here now.
Neill SniderCommissioner of the General Land Office
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile. It also wrongly said this candidate had completed no survey.
An earlier version of this card listed the candidate’s own campaign site, odusfortexas.com, among the pages read in full and listed the section headings seen there; that claim is withdrawn, because when we re-checked on 2026-10-30 the site’s robots.txt asked automated tools, ours among them by name, not to read it, and the copy of its robots.txt we held when it was read has since been replaced, so we cannot show it permitted us then, and nothing from that site is relied on here.
Okey AnyiamJudge, Court of Criminal Appeals, Place 3
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said Ballotpedia records that they did not complete its Candidate Connection survey. That was our error: the copy of the profile we read on 2026-09-20 already recorded the completed survey.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said no campaign site of their own could be found. That claim was wrong, and it was wrong because the search behind it looked only at search-engine results and never at the addresses other trackers already cite.
Scott BristerChief Justice, Fifteenth Court of Appeals
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
Scott K. FieldJustice, Fifteenth Court of Appeals, Place 2
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile. It also wrongly said this candidate had completed no survey.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile. It also wrongly said this candidate had completed no survey.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
Thomas SmithJudge, Court of Criminal Appeals, Place 3
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said the candidate’s own campaign site, timforhouse.com, was read on 2026-09-20 and 2026-09-21 — nine pages reached from its home page, and six of the eight pages its sitemap lists — and that none of them contains cannabis, marijuana, marihuana, hemp, THC, CBD or Compassionate Use; that claim is withdrawn, because when we re-checked on 2026-10-30 the site’s robots.txt asked automated tools, ours among them by name, not to read it, and the copy of its robots.txt we held when it was read has since been replaced, so we cannot show it permitted us then, and nothing from that site is relied on here.
Tom BakerJustice, Fifteenth Court of Appeals, Place 2
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said no campaign site of their own could be found. That claim was wrong, and it was wrong because the search behind it looked only at search-engine results and never at the addresses other trackers already cite.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said Ballotpedia records that they did not complete its Candidate Connection survey. That was our error: the copy of the profile we read on 2026-09-20 already recorded the completed survey.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said their profile carried no Candidate Connection survey, although the same card also said their survey answers were published on that page. That was our error: the survey Ballotpedia shows was completed in 2025, before that sentence was written.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile. It also wrongly said this candidate had completed no survey.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said that two web searches had been run for a campaign site of this candidate’s own, one pairing their name with the office or district and the words "campaign website 2026" and one pairing it with the words "issues platform priorities"; that every result which was not a voter-guide aggregator, a news report or a social-media profile had been opened and inspected by hand; and that no campaign site of their own existed to be had. That claim was wrong to publish, and it was wrong in two ways that are about this guide and not about this candidate: no record of those two searches exists anywhere in this project — not a query, not a result, not the name of an engine, not a note of a failure — while every other class of evidence on this card has one; and the same pair of queries, put to a candidate whose own campaign site this guide has already published a record from, returned nineteen results of which not one was that candidate’s own site. We should not have published a negative about a named person on a search this guide cannot show and has since watched fail a case it should have passed, so the sentence is withdrawn; nothing has been put in its place that claims a site of their own does or does not exist, and the queries, the results and the test that withdrew the sentence are kept at pipeline/data/inbox/campaign-site-search-control-2026-10-12.json.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.
An earlier check here cited the wrong page — a page that does not exist, or a disambiguation page listing several people of that name — and said we had read it. That claim was wrong and has been re-done against the correct profile.
An earlier version of this card said we had searched the Legislative Reference Library’s roll of every member of the Texas Legislature since 1846 for this candidate’s surname, and had read every matching member’s profile. We should not have: lrl.texas.gov asks automated tools not to read the site and names ours among them, that refusal was already saved on our own disk eighteen hours before we began fetching, our robots gate cleared the fetch anyway because it read the rules the site publishes for everyone else and ignored the group that names us, and we took 1,091 pages from the host. That claim was wrong to publish, and it is withdrawn: the worth of a sentence about our own thoroughness is that our method can be relied on, and this one was got by going where we had been asked not to go.