Methodology

The facts on Open America come from government sources. Vote tallies, bill text, sponsorship, committee assignments, floor activity, and lobbying disclosures are all pulled from official public feeds. We do not buy data from vendors, and we do not invent it. If you want to check a figure, this page tells you where it came from and how it was calculated.

We also use AI in a few clearly labeled places, covered in their own section below. It never produces the data or the statistics, only plain-language reading aids.

Data sources

Six official, public sources. Most carry more than one feed.

• congress.gov API (v3). The bulk of the corpus: bills and resolutions, sponsors and cosponsors, amendments, committee assignments, member biographies, CRS plain-language summaries, policy areas, related-bill links, hearings, committee reports, nominations, and the Daily Congressional Record index.
• clerk.house.gov. Three feeds: the FloorSummary XML for live House floor activity, the EVS XML for House roll-call votes with each member's position, and the official member Demographics PDFs (background such as prior military service, medical and legal professions, and freshman status).
• senate.gov. The roll-call vote menu and the per-vote XML for each Senate roll call (per-senator positions, party tallies, question text), the per-day floor-activity XML (the Senate's end-of-day record of business), and committee membership rosters. Some senate.gov endpoints block automated access by network address, so we fetch those through a thin proxy that forwards only to specific senate.gov paths.
• whitehouse.gov OMB. Statements of Administration Policy. The White House's formal position on a pending bill, shown on that bill's page.
• govinfo.gov. The Daily Congressional Record archive, which is the full floor transcript.
• lda.gov. Federal Lobbying Disclosure Act filings: who is lobbying, for which client, on which bills, and on what issues. We link the bills named in a filing back to their pages, and a filing back to the bills it names.

Where AI is and isn't used

No number on Open America is produced by a model. Vote counts, party-unity percentages, and every other statistic are computed directly from the government feeds. AI is used in three places, each labeled where it appears:

• Bill summaries. For some bills we generate a short plain-language summary. The model is given only the bill's own title, status, and official summary, and is told to summarize that and add nothing else.
• Ask about a bill. An optional question box on bill pages, answered from that bill's text. It runs only when you ask.
• Constituent letters. An optional tool that drafts a letter for you to edit and send. It runs only when you start it.

If an AI summary or answer ever disagrees with the bill, the bill is right.

How often the data refreshes

Different signals refresh at different speeds. We try to match each upstream's own pace without hammering it. All times Eastern.

• Every 60 seconds while a chamber is in session for the breaking signals: House roll-call votes, Senate roll-call votes, the House floor summary, and new OMB statements. New votes and floor actions usually appear within a minute or two of being published upstream.
• Every 5 minutes while in session for the full pipeline: newly introduced bills, committee meetings, hearings, and reports. The same pass checks for brand-new bills just past the current high-water mark for each bill type, so a freshly numbered bill is found quickly.
• Hourly through the evening (about 2 p.m. to midnight Eastern) for the Senate's per-day floor record, which the Senate publishes after it gavels out.
• Once a day (around 2 a.m. Eastern) for lobbying filings.
• Both live passes check whether either chamber is actually in session first, so weekend, recess, and overnight minutes cost almost nothing.
• The live floor page receives new actions as a push stream while the tab is open, and falls back to a 60-second poll if the stream drops. New actions appear without a reload.

How every statistic is calculated

One definition per metric. The same source on every page where the metric appears.

Bills introduced (per legislator)

Count of bills and resolutions where the legislator is the sponsor, excluding amendments. Counts span the legislator's full congressional career, not just the current Congress.

Amendments are excluded so the count matches the list you see when you click in. They are tracked on their own and shown with the bill they amend.

Bills cosponsored (per legislator)

Count of bills where the legislator is a cosponsor, with the same amendment exclusion.

Party unity (per legislator)

Of all the legislator's recorded Yea and Nay positions, the share where they voted with their party's majority on that specific roll call. "Their party's majority" is the side (Yea or Nay) that drew more votes from the legislator's own party on that vote.

Independents are compared to the independent caucus rather than to D or R. Present and Not Voting positions are left out of the denominator. If a party splits evenly on a vote, that vote has no majority and does not count either way.

Missed votes (per legislator)

Share of roll calls during the legislator's service where they were recorded as Not Voting, out of the total roll calls during their service. Legislators who joined mid-Congress are not penalized for votes held before they were sworn in.

Vote tallies (Yea, Nay) on a roll call

The headline tally on a vote page is the official count from the clerk.house.gov or senate.gov XML. We separately store every individual member's position. The two reconcile exactly on every recent Senate vote we checked (verified 2026-06-05 across the most recent 50 Senate roll calls, zero discrepancies). If they ever disagree, the official chamber tally is canonical and the per-member data is treated as suspect.

Moving in Congress right now

Bills with floor activity or formal legislative action in the last 48 hours, ranked by most recent action first, then by how much activity they saw. It is cached for 5 minutes on the server. This is a live signal of what the chambers are doing, not a measure of popularity. Where the site shows what is "most watched," that reflects what Open America's own users have added to their watchlists recently, not an editorial pick.

Coverage

Bills and resolutions go back to the 94th Congress (1975). Recorded roll-call votes go back to the 101st Congress (1989). Coverage is deepest for the most recent Congresses, where every feed above is running live; older years are present but thinner. Lobbying filings cover recent years, with the heaviest coverage from 2023 onward.

What we don't have

The current gaps:

• No real-time Senate floor narrative. The Senate does not publish a minute-by-minute feed the way clerk.house.gov does. We ingest the Senate's per-day floor record after the chamber gavels out, and we show Senate roll-call votes within a few minutes of close. For a vote in progress we link to the senate.gov live stream.
• No campaign finance. We track federal lobbying disclosures (who lobbied on what), but not FEC contributions, donor data, or spending. OpenSecrets and FollowTheMoney do that well, and we link to them where it helps.
• No bills before 1975 or votes before 1989. Earlier records exist on congress.gov but are not in our ingestion yet.
• No state, local, or tribal legislatures. Open America covers the U.S. Congress only.

Errors and corrections

If a number on Open America disagrees with its upstream source (congress.gov, clerk.house.gov, senate.gov, lda.gov), the upstream is right and we have a bug. Report it with the page URL and the specific figure. Data correctness is a top priority.

Citation

Open America's data is free to cite, link, embed, and use for research, journalism, classroom work, and civic projects, paid or unpaid. We ask for two things.

• Attribute it as "Open America (openamerica.io)" and link back when the format allows.
• For figures tied to a specific moment (live vote counts, as-of stats), include the date and time of your snapshot. Those change minute to minute during sessions.

We don't sell the data, run ads, or paywall this page. Journalists and researchers who want help citing or pulling something specific can write to hello@openamerica.io and we will respond.