Facts only · No opinions · No spin

Data synced October 24, 2026 · Latest roll call September 30 · Election Day is Tuesday, November 3

Methodology

Methodology & Corrections

How ClearVoting gets its facts, what AI writes, the rules it follows, and what we have corrected. Every count on this page is taken from our data when the page is served.

Where the data comes from

Members, bills and House roll calls come from the Congress.gov API. Senate roll calls come from the Senate's XML records on Senate.gov. Campaign-finance totals come from the Federal Election Commission (FEC.gov). ZIP codes and Senate seats come from two public files, described under ZIP codes and districts and Seats on the ballot.

An automated sync collects the data once a week.

How fresh it is

Roll calls and bills are synced weekly; the edition line on every page shows the last sync and the newest roll call.

Every block written by AI says when the plain-English summaries last changed.

What we cover

ClearVoting holds 1,658 bills — every bill that has had a roll-call vote since January 2021 in the House or Senate. Bills that never reached a roll-call vote are not in our data.

Senate records on member pages include votes on bills, amendments and motions; the 1,477 Senate votes on nominations since January 2021 aren't included yet. House records include every House roll call.

What AI writes, and what it never touches

An AI model writes each bill's one-line summary and its list of provisions, from the bill's title, official summary and a text excerpt. It never touches votes, tallies, dates, names, outcomes or party.

Our rules for neutrality

The AI is given these rules, word for word, every time it writes a summary:

  1. NO adjectives (no "sweeping", "controversial", "landmark", "modest", etc.)
  2. NO value judgments (no "this would help/hurt", "beneficial", "harmful")
  3. NO characterization of intent (no "aims to", "seeks to" — just state what the bill does)
  4. NO political framing (no "progressive", "conservative", "bipartisan effort")
  5. NO editorial characterizations (no "generally seen as", "made it easier/harder", "widely considered"). State only what the rule does, not how people feel about it.
  6. ONLY state mechanisms: what changes, what numbers change, what rules are created or removed
  7. Write at a 7th-8th grade reading level. Use short, common words. Say "cuts" not "rescinds", "makes illegal" not "classifies as Schedule I controlled substances", "stops" not "eliminates the provision", "lets" not "authorizes", "non-citizens" not "non-U.S. nationals", "break-ins" not "burglary" or "larceny"
  8. NO policy jargon or legislative language. Replace terms like "appropriations", "fiscal year", "earmarks", "provisions", "amendments", "authorizes", "mandates", "directs", "enacts", "nationals", "reclassified", "eligibility" with plain alternatives. Say "the year starting October 2025" not "fiscal year 2026".
  9. NO technical jargon without plain explanation. If a concept requires a technical term (like a government agency name), keep the surrounding sentence short and simple. Never stack multiple technical terms in one sentence.
  10. Include specific numbers, dollar amounts, dates, and thresholds — but explain what they mean
  11. When a bill changes a number (fee, cap, limit, threshold, tax rate), include both the old and new values so the reader can see the difference. When a bill affects a group of people, include how many people are in that group if the number is publicly known. Apply this consistently across all bills — the goal is to give readers enough facts to form their own opinion without needing to research further.
  12. Write like you're explaining it to a neighbor, not to a lawyer
  13. When a bill repeals, cancels, or overturns a previous rule, describe the RESULT — not the chain of actions. Instead of 'Cancel a rule that ended automatic approvals for bank mergers', say 'Bring back automatic approvals for some bank mergers'. Avoid stacking negatives. The reader should immediately understand what changes.

Our rules for accuracy

  1. MATCH THE SCOPE EXACTLY. If a bill blocks, affects, or changes ONE specific thing — a single arms sale, one rule, one agency, one program — say exactly that. Do NOT inflate it into a blanket or sweeping action. Write "Block one proposed sale of bombs to Israel", never "Stop selling weapons to Israel". A narrow action described as a broad one is a factual error.
  2. NAME THE BILL'S SPECIFIC SUBJECT. Lead with the actual thing the bill is about — the commodity, agency, country, rule, or program — taken from the title and text. Do not swap a specific subject for a generic category. Say "critical minerals and rare earth elements", not "energy resources"; say "the CPSC", not "federal regulators", when the text names one.
  3. USE ONLY WHAT IS IN THE SOURCE. Build the summary only from the title, official summary, and bill text provided. Do NOT add country names, counts, dollar figures, dates, or other specifics from your own knowledge — they may be wrong or may change over time. If the source does not say it, do not state it.
  4. SUMMARIZE THE OPERATIVE VERSION. The official summary often describes an EARLIER version (as introduced, or as passed by one chamber). If the bill became law or was amended, describe what the FINAL text does — provisions are frequently stripped or changed before passage. If you cannot confirm a provision survived into the operative version, do not assert it. (Example failure: a continuing resolution's debt-limit suspension was removed by the Senate before enactment, yet the introduced-version summary still listed it.)
  5. VARY YOUR SENTENCE OPENINGS. Do not start multiple provision bullets with the same word or phrase (e.g., three bullets all beginning "Removes updated..."). Rephrase so the list reads naturally.

1,528 of 1,658 bills have a plain-English summary. 2 have none yet, and 128 summaries are withheld because they did not pass our automated accuracy check; those bills show their official title. Plain-English summaries were last updated October 24, 2026.

177 summaries were spot-checked by a second AI review.

Member summaries are withdrawn until they can be rewritten without stance labels.

We do not publish arguments for or against a bill. An earlier version showed AI-drafted 'supporters say / critics say' bullets; they were unattributed and have been removed until they can be sourced.

How “Recent votes” are chosen

The “Recent votes” list on each member page is chosen by one rule, and the rule never looks at how the member voted:

  1. We consider the member's roll-call votes on a bill that has a plain-English summary we show and a policy area, where the vote was on the bill itself — a final vote to pass or adopt it, to accept the other chamber's changes, to override a veto or to ratify a treaty — and the member voted yes (Yea or Aye) or no (Nay or No).
  2. Going from newest to oldest, we keep the first such vote on each bill. A bill counts as used only once one of its votes is picked, so a vote the member missed never hides one they cast.
  3. We take one vote per policy area in turn, starting with the area whose newest vote is most recent, and go round the areas again until we have eight. So the first pick is always the newest qualifying vote.
  4. The eight are shown newest first. A member with fewer qualifying votes shows as many as qualify.

Amendments, procedural motions and votes on nominations are never in this list. Every vote in our record is in the member's full record below it; Senate votes on nominations are not in our record yet (see What we cover).

Counts, never grades

We show roll calls voted as a count, never as a grade, and never rank members by it.

A member page says how many of the roll calls in our record the member voted in, and since when. It never turns that into a percentage, a colour or a score.

ZIP codes and districts

To find your representatives from a ZIP code, we use the Census Bureau's table of which congressional districts each ZIP code area overlaps.

Source: U.S. Census Bureau 2020 ZCTA to 119th Congressional District relationship files (Public domain (17 U.S.C. §105)), retrieved September 11, 2026.

https://www2.census.gov/geo/docs/maps-data/data/rel2020/cd-sld/

Shares are land area, not population; PO-box-only and business ZIPs are not in the Census file; a ZIP that crosses a state line is shown with a note.

A district counts as a match when it covers at least 5 % of the ZIP's land area. When one district matches and at least 95 % of the ZIP's land is in California, Florida, New York or Texas, we take you straight to that district. When two to four districts match, we show each one with its share and you choose. When less than 95 % of the ZIP's land is in our four states, we show the match with a note and remember it only if you choose it.

Not sure? Look up your district by street address on house.gov →

Seats on the ballot

Every House seat is on the ballot on November 3, 2026. Of the eight senators we cover, the Texas seat held by John Cornyn is up for its regular election and the Florida seat held by Ashley Moody is up in a special election. Source: senate.gov class lists, retrieved September 19, 2026.

https://www.senate.gov/senators/Class_II.htm · https://www.senate.gov/senators/Class_III.htm

These are facts about the seats, not about the people in them: our data does not say who is running.

Corrections

No corrections logged yet.

Report an error

If something on ClearVoting is wrong, tell us. Use Report an error on this page at the foot of any page, or the button below. A person reads every message, and corrections are listed above with their date.

Send feedback