Has AI Already Read Your Books? Introducing Recall

Share
Has AI Already Read Your Books? Introducing Recall
Introducing Recall

Since 2023, publishers have known that millions of pirated books were copied into datasets used to train AI models. Most publishers still can’t say which of their own titles were among them. Finding out has meant digging through court exhibits, spreadsheets, and one-off lookup tools, one title at a time.

Recall gives publishers and authors a direct answer, with a source for every result. You can search any book, or an entire catalog, against the pirate libraries and datasets that AI companies are known to have downloaded, and see exactly where your books appear.

What Recall checks

Recall searches Books3, LibGen, Z-Library, and Anna’s Archive, the collections named in court filings and public reporting about how AI companies gathered training material. The index currently covers 16.8 million ISBNs and records tied to five AI companies, and it was last updated on October 5, 2026.

We check every edition of a title, so the hardcover, paperback, and ebook are all included even when they carry different ISBNs. When we find a match, you see that collection’s own record of your book, and every claim links to the court record or document behind it. You can read the evidence yourself rather than taking our word for it.

For publishers: check the whole catalog at once

Most publishers manage thousands of titles, often across several imprints. You can upload a CSV, Excel, or ONIX file and check your full list in one pass, using the metadata you already maintain.

The results give your team something to work from. Rights and contracts staff can see which titles were exposed and where. Licensing teams can bring specific titles to conversations with AI companies, in place of general concerns. Editorial and author-relations staff can answer the authors who are already asking whether their books were taken.

For authors: look up your own books

Authors can search by ISBN or by title and author name. If your book appears, you’ll see which collection holds it and the document that shows it. If you have several books or editions, check each one.

A result is information to bring to your agent, publisher, or rights adviser. It can help you understand the AI provisions in your contract and decide what you want to say about future use of your work.

What Recall won’t claim

Recall reports where a book appears. It never estimates whether a particular AI model was trained on a particular book, because no public record can establish that with certainty. A “not found” result covers only the collections we searched, and Recall is not legal advice.

We built it this way on purpose. Publishing has seen plenty of AI tools that offer confident scores with little behind them. A result you can trace to a court document is something you can bring into a negotiation or a contract discussion.

From finding out to setting terms

Recall shows you what was taken. Registering with Amlet lets you set terms for what comes next. Each registered work receives an ISCC identifier, a fingerprint generated from the content itself under the ISO 24138 standard, and you declare your terms for AI training in a form AI companies can read.

Many publishers will use Recall to see where they stand, then register their catalogs so their terms travel with their books from here on.

Check a title

Search a book or upload your catalog at recall.amlet.ai. We offer a monitoring service for publishers – get in touch @hello@amlet.ai