Explore the Epstein Files

A public-records research assistant for the U.S. Department of Justice documents released in the Jeffrey Epstein and Ghislaine Maxwell cases. Ask a question in plain English and get answers grounded in court filings, with direct links to the original PDFs on justice.gov.

2.2M+

Documents

U.S. DOJ

Source

Court records & emails

Format

Topic explorers

Major events, court cases, and document collections.

Methodology

How retrieval, citations, and limitations work.

Resources

Government repositories, journalism, books, and films.

Glossary

Plain-language definitions of legal terms.

Type a question below to start exploring.

About this project

Epstein Files is a free, non-commercial research tool that makes the public document record of the Jeffrey Epstein case searchable through natural language. The corpus combines court filings, depositions, exhibits, and correspondence released by the U.S. Department of Justice, the U.S. House Committee on Oversight and Government Reform, and the Southern District of New York. None of these materials are private or leaked; every cited document is already public.

The interface uses retrieval-augmented generation (RAG) — when you ask a question, the system retrieves the most relevant document passages from a Pinecone vector index and asks a large language model to summarize them. Every claim in an answer is tied to a citation in the form [EFTA00000000] that links back to the original PDF on justice.gov. If a citation isn’t there, the statement isn’t in the documents.

What you can find here

The dataset includes the major events and document collections produced during two decades of state, federal, and congressional proceedings. The topic guides below summarize each one and link to the underlying records.

See the full list of topic explorers for additional collections, including civil cases, flight logs and travel records, and the dataset and corpus documentation.

How to use the chat

Ask specific questions. Vague prompts retrieve scattered passages and produce vague answers; targeted questions retrieve focused passages and produce specific, citable answers. Good prompts name a person, document type, time window, or event — for example, “What charges did the 2019 SDNY indictment include?” or “Who testified at the Maxwell trial about flight log entries?”

When in doubt, click the citations. Each EFTA identifier is a real DOJ document number that resolves to a justice.gov page. If a passage looks important, read the original filing rather than relying only on the summary. The methodology pages explain how retrieval and citation injection work, and the limitations page documents the known failure modes — including hallucinated citations, missing context, and gaps in OCR quality.

Scope and disclaimer

This tool is for educational and journalistic research. It does not constitute legal advice, accusations, or conclusions about any named individual. Many people mentioned in these documents — including in witness lists, contact directories, and flight manifests — are not accused of any crime. Inclusion in a document is not evidence of wrongdoing, and the system surfaces what the documents say without endorsing any inference.

The documents contain references to sexual abuse, trafficking of minors, and other traumatic material. Read with care. If you or someone you know needs help, contact the National Sexual Assault Hotline at 1-800-656-4673 (United States).

Read more about the project on the About page, common questions on the FAQ, and data practices in the privacy policy.