How our ratings work.

For sitting senators and representatives, our ratings measure how their actual voting record compares to what most Americans say they want — across four pillars: Environment, Healthcare, Education, and Immigration.

We're not measuring whether they're “good” or “bad.” We're measuring how often their votes line up with the majority position on issues where most Americans agree (typically 60%+ support across multiple polls).

Voting records come from Congress.gov, the official congressional database. Polling data comes from a mix of well-known pollsters across the methodological spectrum: Pew, Gallup, KFF, AP-NORC, Rasmussen, Trafalgar, and others. We use multiple sources for every claim so no single pollster's house effects drive our ratings.

Cross-ideology scorecards come directly from organizations like Heritage Action, ACU, LCV, and ADA — published by them, not modified by us.

This is the most important question we face, and our answer is built into the design of the system.

We don't decide where a senator stands. Heritage Action does, ACU does, the League of Conservation Voters does, and Americans for Democratic Action does. Their scorecards are visible right on every senator's rating page. When organizations from across the ideological spectrum agree, it's not bias — it's convergence.

We also break out our polling by party affiliation. When 66% of Republicans support drug price negotiation, that's a Republican-majority position too — and a senator voting against it is voting against their own party's voters, not just against ours.

We don't pretend we can. Challengers don't have a voting record, so the Majority Alignment Score, the Word-vote gap, and the Voter gap don't apply to them. We say so clearly on their rating pages.

For challengers, we show their self-described positions on the four pillars, sourced directly from their campaign materials and public statements (with links). Their position on the political spectrum is shown with a dashed marker instead of a solid one — the visual cue that this is what they say they believe, not what they've voted.

When a challenger wins office and starts voting, their rating transitions to the full incumbent system within their first 6 months. We then explicitly compare their early votes to their original campaign statements, generating a Word-vote gap from day one.

It measures the distance between what a senator says publicly and how they actually vote. Some senators have a remarkably consistent record — what they say is what they do. Others say one thing in press releases and vote the opposite way on the floor.

The gap shows up issue by issue, with the exact quote (sourced and dated) next to the exact roll call vote (also sourced and dated). You can see the contradictions yourself.

The Voter gap measures the distance between what a senator's constituents want and how the senator votes. A senator who represents their state well has a narrow Voter gap. A senator who consistently votes against majority opinion in their own state has a wide one.

This is different from the Word-vote gap because constituents may want something the senator never publicly addressed at all.

Voting records update hourly when Congress is in session. The Majority Alignment Score recalculates weekly (Sunday mornings). Cross-ideology scorecards update monthly as organizations publish new versions. Major events (like a candidate withdrawing or a primary result) trigger real-time alerts for members.

Some data sources publish on their own schedule. For example, the American Conservative Union publishes its annual scorecard typically in early spring of the following year. If the current Congress hasn't been scored yet, we calculate an estimate based on the senator's lifetime ACU score and the directional pattern of their recent votes.

Estimated scores are clearly labeled. The estimated score gets replaced with the official score the moment the source organization publishes it.

Yes — for scale, not for opinion. AI helps us match candidate statements to specific votes (it would take a human team months to do this manually for 535 members of Congress on dozens of issues). AI also drafts the initial summary text on race pages.

But AI never makes the ideological judgment. The alignment status (Aligned, Partial, Mismatch) is reviewed by humans, and every AI-generated piece of content carries a clear “AI-generated” badge until a human verifies it. We disclose this throughout the site.

Tell us. Every rating page has a “Report an issue” link. We review every report within 48 hours and respond, whether or not we end up making a change.

We treat reader corrections as a feature, not a nuisance. No rating system is bias-proof, but a system that responds to corrections gets closer to honest over time. You can also email corrections@themajority.us.