About the Data

iRate is published by Dallas AI Metrics, LLC. This page explains who produces the analysis in this section, how it is made, and how to tell us when we get something wrong.

What iRate is

iRate is a public scoreboard for American politics. Visitors rate sitting federal and state officials on a nine-point emoji scale, and those ratings aggregate into a running public verdict for each politician.

Alongside the ratings, each politician page carries their legislative record drawn from official sources — sponsored bills, recorded votes, and committee assignments. The analysis section takes that same underlying data and asks questions of it across the whole of Congress rather than one member at a time.

Who publishes it

Dallas AI Metrics, LLC is the operator of iRate and the publisher of this analysis. The same entity is named in the site's privacy policy and terms of service.

We are not affiliated with any political party, campaign, committee, advocacy organisation, or government body. No one pays for coverage, placement, or a particular framing, and no figure in this section is adjusted for anyone's benefit.

How the analysis is produced

The pipeline is deliberately simple, and every stage is automated and repeatable.

• Collection. Rosters, bills, committee assignments and terms come from api.congress.gov. Roll-call positions come directly from senate.gov and clerk.house.gov.
• Validation. Each refreshed dataset is compared against the previous known-good version. A suspicious change — a sharp drop in record count, a malformed file — causes the previous version to be restored rather than published.
• Computation. A single module computes every statistic in this section from the corpus. There is no manual step where a number could be typed in, adjusted, or overridden.
• Publication. Written commentary is composed separately from the figures. The prose describes what a measure means and how to read it; every specific number and every named member is generated from current data at build time.

That last point is the important one. It means the commentary cannot drift out of sync with the figures as the data refreshes — a hand-written claim about who ranks where would become false the moment the rankings moved, so we do not write them.

Editorial approach

Three commitments govern what appears here.

Descriptive, not causal. These are measurements of what happened. We do not infer motive, and we say so on pages where the temptation is strongest.

Limitations stated on the page. Where a measure is weak, confounded, or easy to misread, the page reporting it says so directly rather than burying the qualification. Several pages in this section conclude that the honest answer is "this does not show what you might expect."

No composite scores. We do not publish an overall effectiveness rating, an ideology placement, or a single number summarising a member. Each would require weighting choices that embed a political judgment while appearing objective.

Use of AI

The written commentary in this section is drafted with AI assistance and reviewed before publication. The figures are not: they are computed directly from official records by code, and no language model generates, estimates, or fills in any number that appears here.

We state this because readers are entitled to know how what they are reading was produced.

How often this updates

The underlying data is refreshed on a recurring schedule, and the analysis is recomputed whenever it changes. Each page states the date its figures were computed, at the top, next to the byline.

Different parts of the pipeline move at different speeds, which is worth knowing if a figure looks stale.

• Roll-call votes appear quickly. Both chambers publish position-level records shortly after a vote is taken.
• Bill status follows within a day or so of an official action being recorded.
• Subject classifications lag furthest behind. These come from Congressional Research Service summaries, which are produced on their own schedule — recently introduced bills often have no classification for some weeks, and are excluded from subject counts until one arrives rather than being assigned a guess.
• Committee assignments and rosters change irregularly, following elections, resignations, and mid-session reassignments.

Because figures move, a number quoted from this section should be cited with the date shown on the page. The same page will legitimately show a different value later.

Corrections

We would rather fix an error than defend it. If something here is wrong — a figure that does not match the official record, a member described incorrectly, a methodological choice that produces a misleading result — tell us.

Contact irateusa@gmail.com with the page, the figure, and what you believe the correct value to be.

Errors in our processing are corrected on the next rebuild. Errors originating in a government source are reported upstream, and we note the discrepancy on the affected page where we are able to.

For the full technical detail of how each figure is computed, see the methodology.