Sleuth Report: Ellie Leonard explains how to trace Epstein FBI documents

Snapshot

---

Important points and key takeaways

Ellie Leonard’s original tutorial shares a research workflow rather than a completed investigation. Her accompanying note says she is preparing a written explanation. The recording demonstrates how dates and identifiers can help assemble related records across a fragmented public collection.

The useful principle is corroboration. A date produces candidates. A production identifier narrows the search. Matching content helps test whether handwritten notes belong with a particular interview. None of these clues should be treated as conclusive alone.

Three distinctions matter: an interview date can differ from the date a report was prepared; a production identifier differs from an EFTA page stamp; and a defense disclosure inventory differs from an inventory of the current public release.

---

Watch Ellie Leonard’s original tutorial

Ellie Leonard’s YouTube recording runs 33 minutes and 18 seconds. The source discussion below refers to the original Substack recording, whose timing differs slightly. The practical steps are an EpsteinWiki research checklist accompanying Leonard’s demonstration.

---

How the demonstrated method works

Leonard’s demonstration at 2:12 through 6:25 starts with handwritten notes, a date and a numbered production index. She searches for the date, identifies a candidate grouping and looks for a corresponding typed interview. She compares content before marking an item as located.

At 7:20 through 11:38, she explores nearby EFTA numbers to locate more notes. This is a discovery technique. It does not prove that every nearby page belongs to the same interview. Her own demonstration encounters a photograph among the records.

At 12:06 through 13:20, a complaint requires a different route: searching its document type and checking its date and subject. At 15:57 through 18:03, she compares another note sequence with a typed interview. A single strategy will not retrieve every document type.

At 29:51 through 30:39, Leonard qualifies her terminology. A numbered group can include witnesses and associated testimony. Her use of the word case is an organizing shorthand, not a verified statement that each group is a separate legal proceeding or one survivor’s complete file.

---

Step by Step Guide

Step One: Start with one record and one question

Choose a public index entry, typed interview or note page. Write down what you want to locate, such as the notes associated with an interview on a particular date. Save the source link, EFTA stamp, PDF page and any other identifier exactly as printed. Keep the untouched download separate from your annotations. Use an existing survivor code rather than trying to identify a redacted person.

Step Two: Read the index columns before searching

The Maxwell index, PDF page 5, provides a safe practice example. Find section 3509. It lists 3509-008 as an interview dated February 27, 2020, and 3509-009 as notes dated the same day. The separate production column says October 11, 2021. Record all three facts. Searching only the production date would send you looking for the wrong event.

Step Three: Search identifiers and dates separately

Start with the exact identifier, then try the document date and type. In a downloaded index, use Find for 3509 or 2020.02.27. If extraction mangles the identifier, inspect the page image instead. For a wider corpus search, try the date in different formats, such as February 27, 2020 and 02/27/2020. Keep a log of which searches returned useful candidates. These are search attempts, not promises that every engine will find the same records.

Step Four: Check the actual pages

Open the PDF rather than relying on a result snippet, automatic summary or handwriting transcription. Check the date, form, page sequence and identifying record numbers. Compare more than one nonidentifying detail before associating notes with a typed report. An FD 302 summarizes an interview; it does not establish that every statement was independently corroborated. Keep a proposed match provisional until the evidence supports it.

Step Five: Explore neighboring pages without inventing a relationship

Nearby EFTA numbers can reveal continuation pages or other candidates. They can also lead to an unrelated photograph, email or document. Check page boundaries, matching headers and content. Do not assume numerical proximity proves an attachment, a common witness or a complete interview set. Likewise, a jump in numbering does not by itself prove a document was withheld.

Step Six: Record the strength of each match

Maintain one row per index item. Include its printed identifier, document date, type, production date, candidate EFTA number, page range, source link and the details supporting your conclusion. Use clear statuses: listed in the index, candidate located, match supported, or not located in the sources searched. If all you have is a matching date, say so. Two copies of the same interview are not two independent witnesses.

Step Seven: Share a reproducible finding

Explain what you searched, what you found and what remains unresolved. Cite both the index and the candidate document with page locations. Preserve uncertainty about illegible handwriting and incomplete sequences. Before sharing screenshots or excerpts, check for exposed names, contact details, medical information and other identifying material. Credit Leonard’s tutorial when using her demonstration, and link the original records so another researcher can repeat the check.

Our FBI file numbers and serial records guide explains the separate task of reconstructing investigative file structure. The 3509 identifiers in this exercise come from a disclosure index; they should not be silently relabeled as FBI case numbers or serials.

---

Evidence appearances and exact references

EFTA00010025, PDF page 5, stamped EFTA00010029 contains section 3509 of a United States v. Ghislaine Maxwell 3500 material index. The mirrored PDF and database extraction were readable. The official DOJ copy was indexed in public search, but a direct browser attempt returned a 404. Direct original verification therefore remains incomplete.

The index pairs interview 3509-001 with notes 3509-002 on September 19, 2019; interview 3509-003 with notes 3509-004 on November 14, 2019; and interview 3509-005 with notes 3509-006 on December 17, 2019. It lists complaint 3509-007 on January 17, 2020, and pairs interview 3509-008 with notes 3509-009 on February 27, 2020. These entries show an October 11, 2021 production date.

This supports the dates and document relationships used in the tutorial. It does not identify the public EFTA location of every interview or authenticate each handwritten page Leonard associates with it.

EFTA00276919, first page corresponds to the later note sequence discussed around 15:57 through 17:44. Its available transcription begins with February 27, 2020, consistent with the index date. The mirrored PDF contains 29 pages. The database’s selected page response exposed only one page, so that response was not used as the full document page count.

Date agreement is a corroborating lead. The complete handwritten sequence and typed interview were not independently compared page by page for this report. Their association remains a provisional match.

EFTA00151669, first page is a starting reference Leonard mentions near 31:37. The database returned OCR text and a PDF reference. No reliable interview date or complete relationship to another record was established here, so no substantive claim is drawn from it.

---

Why this matters

The DOJ Epstein Library warns that handwritten material and other formats may be unsearchable or produce unreliable results. That official warning supports the tutorial’s practical concern: a failed text search cannot establish that a record is absent.

An index can make a search auditable. Researchers can distinguish an item listed in historical discovery from a candidate public document, a confirmed match and an item not yet located. That preserves questions about completeness without prematurely asserting concealment.

EpsteinWiki’s Communications research guide provides related context for working with email, text message and phone records. Leonard’s contribution here is a practical demonstration of cross checking a different set of investigative materials.

The database’s retrieval documentation distinguishes raw OCR from appended transcriptions. A transcription can help locate a page. Researchers should inspect the page image before relying on uncertain handwriting.

---

Uncertainties, open questions and survivor protections

Leonard makes broader claims about missing material, selective redactions and DOJ intent. This review does not verify a numerical total of missing files or establish the motives behind a particular redaction. Differences between versions require examination of provenance, processing and context.

The recording also includes commentary about accounts involving Donald Trump. Those assertions were not independently evaluated against complete interview and trial records for this methods report. They are not adopted as new findings. A person’s appearance in records does not establish wrongdoing.

Can every index entry be tied to a verified public document? Do candidate note sets have the correct start and end pages? Are there revised or duplicate versions? Which transcription ambiguities remain after checking the images? These are verification questions, not gaps to fill by inference.

Survivor privacy comes before reconstructing a file. This report does not reproduce private names, family details, identifying life histories, contact information or exposed note images. Existing anonymized codes should remain in place. A missed redaction is not permission to republish an identity.

Review scope: the original written post and the complete available automatic transcript, from 0:01 through 32:43 of a 32:52 recording, were read. Automatic captions can mishear dates and identifiers. Only the selected documentary checks described above were completed. This is a provisional evidence review, not a certification of the reconstructed file sets.

---

Questions and Answers

Does a matching date prove two documents belong together?

No. It narrows the search. Compare the type, sequence, identifiers and content before treating the relationship as supported.

Does the index list everything currently public?

No. Its title and production column describe material disclosed in the Maxwell case. A listed item still needs to be located and checked in the public collection.

What if a search finds nothing?

Record the source, query and date searched, then try another identifier or inspect the scan. Handwriting and OCR errors can defeat text search. Use “not located in the sources searched,” rather than claiming the record does not exist.

Should I publish an identity exposed in an unredacted copy?

No. Keep survivor codes and remove identifying details from your shared research. An accidental disclosure does not remove the need to protect the person.

---

Sources

• Ellie Leonard, How to find Epstein FBI Documents: original public recording and full available transcript, published October 29, 2026, in America/Denver.
• EFTA00010025: Maxwell 3500 material index, PDF page 5, stamped EFTA00010029.
• Official DOJ index PDF: indexed original location; direct browser access returned a 404 during this review.
• EFTA00276919: first page transcription and 29 page mirrored note sequence. Its full association with the typed interview remains provisional.
• EFTA00151669: starting reference mentioned in the recording, checked for record existence only.
• DOJ Epstein Library: official search limitation and privacy notices.
• Epstein Files Research Database documentation: retrieval methods and the distinction between OCR and transcriptions.
• EpsteinWiki Communications research guide: verified internal research context.
• Ellie Leonard’s YouTube tutorial: original YouTube tutorial.
• EpsteinWiki: FBI file numbers and serial records: related guide.