PayAuthorscom

Are your books in pirate libraries?

Check your name free against the dated catalogs courts found Anthropic and Meta copied.

We match your books and articles to the dated pirate-library catalogs, down to the file records they list, and to the timeline of when each AI company copied those libraries, then to your U.S. copyright registrations. The name check is free.

Authors who opted out of the Anthropic settlement, and authors whose books were never on its Works List, kept their claims, and many are bringing their own suits. Courts and the companies’ own filings show AI companies copied whole pirate libraries. A PayAuthors report gives you and your lawyer the file records, dates and registrations to start from.

See the timeline of the AI copyright cases →

File-level records: the library’s ID, plus the MD5 fingerprint and torrent piece where they are recorded →

See a sample reportA live Registration Details report for Andrea Bartz, named plaintiff in Bartz v. Anthropic: 51 records, 15 titles.

Have a short list? Check up to 10 author names or ISBNs a day from a spreadsheet. For a longer list, ask for a quote.

Publisher, university press or estate? Check your registered books by rightsholder name.

Universities, libraries, publishers and firms: ask for a quote

What the check searches

  • LibGen 25 dated catalog snapshots, 2015–2025, and their torrents
  • Z-Library PiLiMi (2022) and later catalogs to 2025, via Anna’s Archive (metadata only; no book files)
  • Books3 the 2020 file list
  • Torrent pieces 9.1 million LibGen and PiLiMi files located to the exact piece
  • U.S. Copyright Office 8 million registrations, 1978–2025
  • Anthropic Works List 482,420 entries from the settlement’s public lookup
  • Open Library format and edition for 2.6 million ISBNs
  • Internet Archive nonprofit lending library, not a pirate site
  • Sci-Hub and LibGen scimag journal articles: separate check

The data behind each check

Updates with each data build · last built Oct. 5, 2026 (payauthors-2026-10-01-m64) · How we count

Catalog records
22.4MLibGen, Z-Library and Books3, plus the Internet Archive’s nonprofit lending library
Dated catalogs
31LibGen 25 · Z-Library 4 · Books3 1 · Internet Archive 1, Nov. 2015 – Oct. 2026
U.S. copyright registrations
8.03Mpublished and unpublished literary works, 1978–2025
Registrations matched
1.52Mto catalog records, each with its method
Files located in torrents
6.95MLibGen files, each with its ID, MD5 and piece layout · what that means
Journal-article DOIs
85.4Min Sci-Hub and LibGen scimag

File-level records: library ID, MD5 fingerprint and torrent piece

Every LibGen and Z-Library line in a report names the exact file the library offered, by the library’s own ID, with its MD5 checksum where the catalog records one and its torrent and piece layout where that torrent is archived. Not every line has all three; the report shows which it has. It is the record a lawyer compares with a company’s file lists in discovery.

1

The library’s own ID

The number LibGen or Z-Library gave that one file, so the file is named the way the library itself names it.

Z-Library ID 5276476
2

The MD5 fingerprint

The catalog-reported MD5 checksum: the library’s own record of a checksum of the file’s exact bytes. We hold no files and have not recomputed it. A copy produced in discovery can be checked against it byte for byte, without anyone opening the book.

MD5 721630f8116455be2e24f39246759772
3

The torrent and its pieces

The BitTorrent bundle that carried the file, and how the file sits in that torrent’s pieces: its size, the piece length, how many pieces it spans, and the largest share of the file inside one piece. A .torrent file holds file names, sizes and checksums, which you need before you can download the content. It isn’t the content.

pilimi-zlib-5240000-5329999 · file 320 KiB, inside one 4.0 MiB piece

Many books sit wholly, or mostly, inside one torrent piece

BitTorrent checks data piece by piece. When a book’s file lies inside one piece, every verified copy of that piece holds the whole book, which matters when a company says its torrent uploads carried only fragments. Whether one uploader sent all of a piece has to come from that uploader’s records.

The example above is from the sample report: The Herd by Andrea Bartz, as listed in the Pirate Library Mirror (PiLiMi) index of Z-Library. The same book’s LibGen fiction file (ID 2212710) spans two pieces, with 62% of it in one.

What courts have found about the downloads, and what is still open

Summary-judgment orders state that Anthropic downloaded Books3 (early 2021), at least five million LibGen books (June 2021) and two million from the Pirate Library Mirror (July 2022) (Bartz; Anthropic’s own interrogatory answer gives the days: Feb. 25, 2021; June 8–19, 2021; July 8 to Aug. 10, 2022), and that Meta first used LibGen in October 2022 and downloaded Anna’s Archive in early 2024 (Kadrey). Books left off Anthropic’s settlement list, and books whose owners opted out, kept their claims (left out?). An author still has to show ownership, a valid registration, and that the book was in the library when it was copied, and courts weigh defenses such as fair use. Our data speaks to registration and timing.

Each mark carries its source. The full timeline adds companies’ own papers and the allegations, and compares every download with the snapshot just before it.The lawsuits, company by companyOpen the full timeline →

What the report covers

Every record under your name, edition by edition. Metadata only: no book contents are held anywhere.

LibGen, 25 snapshots

Non-fiction and fiction catalogs, Nov. 2015 – June 2025. Each edition with the snapshots that list it.

Z-Library, 2022–2025

The PiLiMi index (July and Sept. 2022) and later catalogs to Jan. 2025, with dates added.

Books3

Matching names in the 196,608-file list, and whether an edition’s ISBN was found in the Books3 texts.

Internet Archive lending

Scanned print books in the nonprofit lending library, with the date each was added. It is not a pirate site. It is listed because pending AI cases allege that AI companies obtained its lending books; no court has found that.

U.S. registrations

Candidate TX matches, 1978–2025, with the match method, the claimant and a link to the Office’s record.

The file record

The file record each catalog lists: catalog number, format and size, the MD5 checksum where the catalog records one, and the torrent that carried it where that torrent is archived: the identifiers the libraries themselves use. Every source we searched is logged by SHA-256.

No black-box conclusions. Sources, rules, dates, identifiers, and limitations are disclosed in the Methodology.

What a result looks like

One line from the sample report, as a Registration Details report shows it: the catalog record, its dates, the file record, the torrent that carried it, the registration and who claimed it.

In plain terms: The Lost Night was listed in LibGen from Feb. 2020 to June 2025, including the June 2021 catalog, the month a court found Anthropic downloaded LibGen. It was registered Nov. 7, 2019, before that download.

Evidence · Andrea Bartzpayauthors-2026-10-01-m64

The Lost Night

LibGen fiction ×1PiLiMi ×1Z-Library ×2Books3 ×1Registration match§ 412 timingISBN in Books3
LibGen fictionBartz, Andrea2019 · English · Crown/Archetype · ISBN 9780525574712
in 17 of 25 LibGen catalog snapshots: first Feb. 2020, last June 3, 2025 · added to the catalog June 12, 2019 (LibGen’s own record)
The technical record (for your lawyer)
  • LibGen fiction ID 2157462 · epub · MD5 8b2a0eab82719cfd5ab89c3e4fd6e3de
  • torrent f_2157000 · file 12 MiB, spans 4 pieces of 4.0 MiB, so no one piece holds the whole file (the largest part in one piece is 33% of the file)
Registration match TX0008858704 (Nov. 7, 2019) · match: ISBN (confirmed by surname or title)
TX0008858704 — THE LOST NIGHT: A Novel.
  • Publication: published Feb. 26, 2019
  • Created: 2019
  • Authorship: Bartz, Andrea — author
  • Claimants: Andrea Bartz Inc., (organization)
  • Recorded documents: none in the Office’s bulk data through June 26, 2025
Passes the date screen (§ 412): Anthropic, Books3 (early 2021); Anthropic, LibGen (June 2021); Meta, LibGen (October 2022) · and 3 more documented events
  • The catalog record as the library listed it: author field, year, publisher, ISBN.
  • The dates: which of our 25 dated snapshots list it, and when the library says it was added.
  • The file: catalog ID and format, with the MD5 checksum where the catalog has one.
  • The torrent: the named torrent that carried the file, and how the file sits in its pieces. When a file lies inside one piece, every verified copy of that piece holds the whole file; whether one uploader sent all of it has to come from that uploader’s records.
  • The registration and how it was matched, with the claimant named on the registration (the owner when it was registered, not necessarily today). The shaded Office record, with any recorded transfers, is in the Registration Details report.
  • Registration date vs. copying date: whether the registration is dated before the download (we date the download as the later of the download start and the library’s own record of when the copy was added), or within three months of a publication that came first. Without that, statutory damages (up to $150,000 per work if the infringement was willful) and attorney’s fees are generally unavailable under 17 U.S.C. § 412, unless an exception applies; the report performs a limited date screen. Where the dates overlap or no download date is known, the report says so and does not count it.

How it works

1

Check free

Type a name or ISBN. You see counts by source and by snapshot, and the titles found, to confirm which are yours.

2

Choose a report

One report, $95: the pirate-library records we find for your books, matched to your Copyright Office registrations.

3

Get your report

Your private report page opens as soon as you order, with PDF, CSV and JSON files. Within 7 days a PayAuthors reviewer checks the matches behind your counts and updates the same link; we e-mail you when the reviewed version is ready. The link stays live until September 1, 2027.

A catalog listing is a statement about that catalog on that date, not proof of any later use of the file. Registration matches are candidates to check against the linked record, and a missing match is not evidence of non-registration. Terms and limits.

Know an author whose books could be in these libraries?Send them the free name check. It takes a minute, needs no account, and shows which dated catalogs list their books.

Questions

Why would I need this?

Authors, their agents and their lawyers often want to know which of an author's works appear, and when, in the pirate libraries named in the AI copyright cases (the LibGen catalogs and Z-Library, through the PiLiMi mirror and later catalogs) and in the Books3 file list; the Internet Archive’s lending library, a nonprofit library and not a pirate site, is listed separately because pending AI cases allege that AI companies obtained its lending books (no court has found that). This report lists the matching records, edition by edition, with candidate U.S. registration matches, in a form you can review with your lawyer or keep with your records. It does not show that any company copied or used a work. Authors left off the Bartz v. Anthropic Works List (no ISBN, late or collective registration, a mismatch) kept their claims; see Left out of the Anthropic settlement?

There are free tools. Why pay?

Use them first. The settlement’s lookup tells you whether a book is on the Bartz Works List, and The Atlantic’s search covers one early-2025 copy of LibGen. Our free check covers 25 dated LibGen catalogs, Z-Library, Books3 and the Internet Archive. The paid report adds the dates of each listing against each documented download, the file records, and your matched registrations with a registration-date screen relevant to whether statutory damages may be available.

Where does the data come from?

Twenty-five Library Genesis catalog snapshots (non-fiction and fiction database dumps) from November 14, 2015 to June 3, 2025, preserved by the Internet Archive; the Pirate Library Mirror index of Z-Library (a metadata-only copy re-hosted on Hugging Face, with its release lists); later Z-Library records from Anna’s Archive’s index (August 2022 to August 2023) and the MajinBook catalogue (January 2025); the Internet Archive’s lending-library catalog, through its public search API; the Books3 file list and ISBN list published by Peter Schoppert (psmedia/Books3Info); and the U.S. Copyright Office's public tabular registration data for non-dramatic literary works, January 1, 1978 through 2025. Only catalog metadata is used; no book contents are held. The report lists each file and its checksum; the Methodology has every source and date.

My name is common. Will I get other people's books?

Possibly. The report lists the records whose author field matches the name you typed; it does not decide which of them are yours. The default "close together" match keeps the parts of your name inside one author name, which removes most co-author list noise. If you publish under a middle name or initial, search that form too — each search is its own report. A report covers one author, so it needs a first name or initial and a surname, unless the author publishes under a single name; records that give only your surname are included and marked so you can leave them out. If your report still includes another person’s books, tell us within 14 days and we will correct it or refund it.

The free check says zero. Is that final?

It means no record in any of these catalogs carries that name as typed. Try other spellings, an initial instead of your first name, or a pen name. Some catalog entries carry no author at all (about 138,000 non-fiction and 40,000 fiction entries, and about 19,000 Books3 files) and cannot be found by name.

Can I get a refund?

If the report is unusable, not as described, or includes books by another person with a similar name, write to support@payauthors.com within 14 days: we will correct it at no charge or refund it, as you prefer. Because the report is issued as soon as you order, we cannot refund a report simply because the results were not what you hoped for. See the refund policy.

What does a match actually show?

That a record or file name carrying the name appears in a dated source: a LibGen catalog snapshot (the report says which ones, from 2015 to 2025), a Z-Library catalog (2022 to 2025), the Books3 file list or the Internet Archive’s lending library. It does not by itself show that any company downloaded or used the work. Read the full explanation.

Do you cover Anna's Archive and journal articles?

Anna's Archive is a search engine and mirror that combines other collections, chiefly LibGen, Z-Library and Internet Archive lending books; this report covers those sources, with dates, using catalog metadata only (no book files). Journal articles are checked separately and free, by ORCID iD, name or DOI, against Sci-Hub's 2017 DOI list and LibGen's scimag database: check your articles.

Is this legal advice?

No. PayAuthors provides litigation support: a data report your lawyer can work from. It is not legal advice. Whether a listing or registration matters for any claim is a question for your lawyer.