ISBNdb Pulls Books-for-AI Sourcing Page After 404 Media Report

A book metadata company quietly pitching itself as a bulk supplier of physical books to AI labs is the kind of small item that tells you where the training-data market has actually gone. 404 Media reported on July 21 that ISBNdb, which describes itself as 'the card catalog of the book world' for more than two decades, had put up a landing page titled 'Printed Books Sourcing for Your AI LLMs Dataset Needs.' By July 30 the page was gone and the company was denying it had ever run the service.

The removed page, in reporting by Samantha Cole, made the pitch surprisingly candidly. It conceded that 'the optics problem is real' and that 'AI company destroys two million books is not a headline that generates sympathy,' while arguing that 'purchasing paper books at scale from the secondary market does not deprive any creator of income they would otherwise have received.' The offer, per 404 Media, was for books 'tailored to your LLM training needs, delivered at the scale AI demands,' with orders reportedly going up to one million books at a time.

ISBNdb's statement walked all of it back. 'ISBNdb has never purchased, scanned, or sold a book,' the company said, characterising the page as 'a test of market interest.' That framing is convenient, but it is also plausible. The more interesting fact is that a company sitting on a book metadata business looked at the AI training market and thought a physical book sourcing service was worth publicly pitching at all.

The honest caveat is that the reporting does not tell you whether any AI company actually bought anything through ISBNdb, or which of its rivals are quietly offering the same service without a public landing page. It also does not resolve what happens to the physical books in that model. Anthropic's documented book scanning practice, revealed in a January 2026 copyright lawsuit filed by authors, involved obtaining and scanning millions of printed books, but ISBNdb's specific plan is now unknowable because the page is gone.

What is worth watching is the market on the other side of this. The more embarrassed intermediaries get about being publicly named, the better the case looks for AI companies that can point to licensed and disclosed training-data supply chains, and the more valuable that supply chain becomes for the publishers who can help build one.