WorkJuly 21, 2026via 404 Media
AI Companies Are Buying Tons of Old Books Because They're Free of AI Slop
Why it matters
AI companies are systematically moving away from internet-scraped data toward curated printed books, revealing a fundamental crisis in training data quality and a new supply-chain vulnerability. This signals that the 'AI slop' problem—models trained on AI-generated content—is forcing expensive operational pivots and creating business model shifts.
Key signals
- ISBNdb positioning old books as 'slop-free' training data sources
- AI companies explicitly buying printed books to avoid AI-generated training contamination
- Market acknowledges 'optics problem' around data sourcing
- Shift from free internet scraping to paid book acquisition model
- Reveals data quality crisis as models train on their own outputs
The hook
NOBODY TALKING: While everyone fights over AI model capabilities, a $2B book-sourcing market just emerged to solve a quieter crisis: AI companies can't trust their own training data anymore.
ISBNdb, a company that sources printed books for AI companies to turn into training data, tells clients “the optics problem is real.”