Metodología

How we choose issues

Each type of office has one fixed list of issues, and every candidate in a race gets the same rows. The lists focus on what each office can actually affect and on the topics Texas candidates are campaigning on in 2026. Each question names a topic, not a preferred answer. The current lists were locked on Oct 20, 2026 (version 1.1) and do not change in the middle of a cycle except to add a topic for every candidate at once.

Judicial candidates are compared on qualifications instead of issue positions, because the Texas Code of Judicial Conduct (Canon 5) bars judicial candidates from making pledges about cases or legal questions that could come before them.

How we write summaries

Each summary describes what a candidate has said or done, in one or two sentences, using the candidate's own framing where possible. We do not characterize positions as moderate, extreme, popular, or unpopular. An automated check blocks loaded words before anything is published.

Quotes are reproduced exactly as spoken or written, limited to 25 words, and set in a different typeface so readers can tell the candidate's words from ours.

"Says" describes a candidate's public statements. "Voting record" describes votes and official actions, and appears only for current or former officeholders. When several candidates cast the same vote, the vote is described with identical wording for each of them; only how each person voted differs.

When we searched and found nothing, the cell says "No public position found" with the date we searched. When a topic has not been researched yet for a race, the cell says "Not yet researched" in a quieter style, so we never imply we looked when we did not.

Source tiers

Every claim has a numbered source. Primary sources are the candidate's own words or official records: campaign sites, filings, vote records, debate transcripts. Reference sources are nonpartisan compilations such as Ballotpedia or Vote Smart. News sources are reporting by named outlets. We prefer primary sources and cite news only to locate a statement or vote. Every source shows the date we retrieved it. We are saving archived copies of every source so links keep working; when one exists, the source panel links to it.

How candidates are ordered

Contests and candidates appear in official Texas ballot order. Offices follow Election Code 52.092. Party order follows Election Code 52.091, which ranks parties by their vote for governor in the last gubernatorial election; independents follow the parties. We never sort by party, polling, or fundraising, and every candidate gets a column of identical width and the same neutral party label. Declared write-in candidates are listed separately because their names are not printed on the ballot.

How we use AI, and how every item is human-reviewed

We use AI research agents to find candidate statements and records across public sources and to draft summaries. Agents may only cite pages they actually opened, and must mark a cell "not found" rather than guess. Nothing an agent writes is published directly: every change is reviewed by a person in a pull request before it reaches the site, and independent verification passes re-open sources to confirm they support each claim. The full history of every change is kept in version control.

How candidates can submit statements

Candidates can submit a statement or a correction to their own entry through the correction form, with a source we can check. When we send questionnaires, every candidate in a race will get the same questions, the same word limit, and the same deadline, and answers will be published exactly as received under "In their own words," separate from our researched grid.

How corrections work

Anyone can report a correction from any page. We review each report against the source provided, usually within 48 hours during early voting. When we change a published position, the previous version stays visible with the date it changed, and the change appears in the log below.

What we don't do

We do not endorse candidates. We do not rate, score, or grade positions. We do not predict outcomes or compute our own polling averages. We do not accept money from candidates, parties, or political committees. We do not store your address: it is sent once to the U.S. Census Bureau geocoder to find your districts and is never saved or logged. Your picks stay on your device.

Issue lists

Recent data updates

1. 20 oct 2026Issue list: Added rows: federal (elections and voting, crime and public safety, transgender policy, Israel and Gaza, religious liberty and faith communities); state (transgender policy, religion and public schools, religious liberty and faith communities, home insurance and housing costs). Selection test: topics Texas candidates campaign on in 2026 and that voters rank as priorities.
2. 19 oct 2026Add build-time OG images, verification schema, research pipeline commands
3. 19 oct 2026Initial build: TheOpenBallot MVP scaffold

Generated from the version history of our data files. Reportar una corrección.