We want the site tested hard. Below are checks with answers you can verify yourself, what to expect when you run them, and how we measure our own accuracy. If a result looks wrong, send it to support@payauthors.com with the name you searched; we look into every reported error, and corrections are logged in the change log.
First, what a result is
A result says that a catalog record or file name under the name you typed appears in one of these sources on a given date: LibGen catalog snapshots from 2015 to 2025, the Z-Library catalogs (PiLiMi and later), the Internet Archive's lending library, and the Books3 file list. It does not show that any company downloaded, copied or trained on that book. Who copied which library, and when, comes from court orders, the companies' own papers and filings, or allegations, and each is labeled with its grade. Registration results are candidate matches to U.S. Copyright Office records, to be checked against the linked Office record.
Five tests with known answers
1. A title and registration number from a court filing
Courts have works lists that pair a title with its registration number. The Bartz v. Anthropic opt-out list (Dkt. 619-3, Ex. J) and the publishers' works lists in Cengage Learning v. Google (No. 1:24-cv-04274, S.D.N.Y., ECF 1236-1) are public. Pick an author from one, run the free check, and see whether the report links that title to the same registration. Our measured rates on these lists are on the Methodology page.
2. A public-domain author
Try Mark Twain or Charles Dickens. You should see a public-domain notice: works published in 1930 or earlier are in the U.S. public domain, so a registration matched to one of their books can cover only what a later edition added (an introduction, notes, a translation, illustrations or an adaptation), and it belongs to whoever created that material.
3. A common name
Try John Smith or Michael Lewis. You should see a notice that several authors share the name, with their birth years from Open Library. The records shown include other people's books. Before a report is bought, the title list lets the buyer untick any book that is not theirs; only the books kept are counted.
4. Your own name
If you have not published a book, the records under your name belong to someone else who shares it, and a report for that name would list their books. The title list shown before buying lets anyone untick books that are not theirs.
5. A book title instead of a name
Type a title. The check offers a title lookup that lists the catalog records for that title with their authors, so you can see whose copies are listed.
Why our numbers differ from other lookups
Other public lookups usually search one copy of one library at one date. We search 31 dated catalog snapshots across several libraries, and a book listed in a 2019 LibGen snapshot but removed before a later one still shows, with the dates it was listed. We combine catalog entries that describe the same edition and show the count. A free check counts records and title lines; a report counts books (one work across its formats and editions) and registrations. Where counts differ, compare at the level of a single title: the report gives each record's library ID and MD5 so it can be found in the source catalog.
How we measure accuracy
Two measurements are published on the Methodology page, each with its date and the rules in force. Recall: how often we link a title to the right registration, tested against works lists filed in court. Precision: how often a match a report counts is the same work, checked by hand by a person who was not involved in writing the matching rules, on matches drawn at random. Only a person's check is used for the precision figure. Software checks help us find errors to fix, but they are never published as accuracy.
Questions reporters ask
Does a match mean an AI company copied this book?
No. It means the book is listed in a library that a court order, a company's own filing or paper, or a complaint says a company downloaded or used. Whether a particular file was among what a company took comes from that company's records. Reports say which grade each event has and give the file identifiers that discovery can test.
Who decides whether a match is right?
Automatic rules compare each catalog record with the Copyright Office record and confirm a match only when every check passes. Matches the automatic rules can’t settle are reviewed by a large language model; a person checks samples, and published accuracy figures come only from a person’s checks. A match nobody has confirmed is shown as Not yet checked and is left out of the counts. See Measured accuracy.
Why does a public-domain author show registrations?
Modern editions of classics are often registered for what they add: an introduction, notes, a translation or illustrations. Those registrations are real, but they do not cover the original text, and they belong to the editor, translator or publisher, not to the original author.
Why might the count include books by other people?
Catalogs list names, not people. Several authors can share a name, and catalogs also list editors, translators and subjects in the author field. The free check shows every record under the name; a report counts only the books the buyer leaves ticked, and labels lines that need checking.
How current is the data?
Each page and report names its dataset build and the date it was built; reports also name the analysis revision. The current build and every change to the rules are listed in the change log.
Can a result be reproduced?
Yes. A report lists the dataset build, each record's library ID and MD5 checksum, the torrent that carried it where known, and each registration number linked to the Office's own record. The sources and their checksums are on the Methodology page, and a published sample of reviewed matches can be rechecked line by line.
How does PayAuthors make money?
The name check is free. Reports are paid: the Registration Details report for a name is $95 and the journal-article report is $35. Law firms license reports and data under written engagements. Prices are on the home page and the law-firm page.
Is this legal advice?
No. PayAuthors provides research. Whether a claim exists, who owns a work today, and any deadline are questions for a lawyer. PayAuthors is not affiliated with, or endorsed by, the U.S. Copyright Office.
Can I get a free report to test?
Journalists can request a free press code for test reports on the press page with a work e-mail address.