Sources and methodology

Where the data comes from

Records are copied from the sources below and refreshed periodically, most recently on Oct. 1, 2026.

Federal Judicial Center

The Biographical Directory of Article III Judges is Benchlight's authority for who sits on the federal bench and how they got there: identity, education, professional career, and every federal judicial appointment with its nomination, committee action, Senate vote, confirmation, commission, chief-judge service, senior status and termination. Benchlight holds it in full — all 4,077 Article III judges since 1789.

CourtListener

CourtListener, a project of the non-profit Free Law Project, is Benchlight's source for the case record: dockets, opinions, panels, oral-argument audio, and links to each court's own documents.

The courts' own websites

Where a court publishes its opinions as PDFs, Benchlight reads the document to establish who wrote it and who sat.

Administrative Office of the U.S. Courts

The judicial vacancy lists are the official record of which seats are open, why, and who has been nominated to them.

Congress.gov

Congress.gov supplies the Senate half of an appointment: the nomination's citation, the date it was received, referral to the Judiciary Committee, hearings, floor action, and the confirmation vote — or its withdrawal or return.

American Bar Association

The Standing Committee on the Federal Judiciary publishes a ratings chart for each Congress. Benchlight records the committee's rating as published, including where a majority and a minority differed.

What Benchlight computes

Anything Benchlight worked out for itself is labelled Computed by Benchlight where it appears. That currently covers:

• Service status — active, chief, senior or former, derived from the FJC's commission, chief-judge, senior-status and termination dates.
• Case status — a readable state such as “Argued — awaiting decision”, derived from filing, argument and decision dates rather than reproducing an upstream code.
• Authorship and panels read from documents — where a court's PDF states who wrote an opinion and who heard the case, the name is matched to a judge in service on that date. Where the name is ambiguous — two judges share a surname, or a visiting judge sits from another court — Benchlight records the name as printed and leaves the seat unmatched rather than choosing between them.
• Panel colleague counts — how often two judges appear on the same panel, counted across the cases Benchlight has ingested. This is a count of shared panels within a window, not a career total, and it says nothing about whether they agreed.
• Writing counts by type — counted from writings attributed to a judge in the ingested records.

What Benchlight does not do

• It does not publish ideological scores or similar ratings.
• It does not guess. Where a panel seat cannot be matched to a judge with confidence, the seat is shown with the name the court published and marked unmatched, rather than being attached to a plausible person. Where an opinion has no recorded author, the page says so.
• It does not merge sources that disagree. Where the vacancy list and the Biographical Directory describe the same seat differently, both are shown as their own source states them.
• It does not republish the corpora it draws on. Judicial documents are linked at their source, not hosted here — with one deliberate exception. A small number of the courts' own documents are set out in full, in their own words: landmark opinions, star-paginated so a passage can be quoted and cited from this page rather than from a PDF, and the Supreme Court's order lists, which dispose of most of what reaches the Court and are replaced by the next list a week later. Those pages name the document the text was read from and link it, and every sentence on them is checked back against it.
• It uses no AI-generated text. Every sentence describing a judge's career, appointment or writings on this site is either a source field or a Benchlight computation of the kind listed above.

What this site collects

Benchlight has no accounts, sets no cookies of its own, and stores nothing about you on its own servers beyond ordinary web-server logs.

It does use Google Analytics, which sets its own cookies and reports the pages you visit to Google. Search terms are deliberately withheld: a search is reported as a visit to /search with the query stripped, because what a reader is looking up here — which judge, which case — is the most revealing thing this site handles, and it is not ours to pass on. Nothing you do on this site is exposed publicly.

Known limitations of this prototype

• Benchlight holds 27,539 cases and 29,009 writings across the Supreme Court and all thirteen courts of appeals. That is a window on each court, not its complete history, and it is much deeper for some courts than others. Judge and appointment records, by contrast, are complete.
• Case histories are built from filing, argument and decision dates rather than from every docket entry, so a case page shows the shape of a case rather than every step in it.
• Panels are ordinarily identified when a decision issues. For an appeal that has been argued and is awaiting decision, the panel is usually not yet on the public record, and the case page says so rather than implying no panel exists.
• Where a court publishes no document Benchlight can read, and no provider records an author, a writing is shown as unattributed. That is a gap in what is known, stated as one.