Workers at the site confirmed that their primary role involves receiving bulk shipments of printed books, cutting off their bindings to facilitate high-speed scanning, and destroying the physical copies once the text is digitized.
This operation matters because AI companies are facing a shortage of high-quality, human-written text to feed their large language models (LLMs).
Books printed before 2022 are particularly valuable because they are guaranteed to be free of "AI slop," or low-quality AI-generated content, which can cause "model collapse"—a technical failure where a model's performance degrades because it was trained on its own synthetic output.
By targeting physical books, Amazon can access vast amounts of data that have not yet been scraped from the internet, helping its "Nova" family of AI models compete with rivals from OpenAI and Google.
The destruction of these books is a strategic and legal mechanism used to accelerate data ingestion while navigating copyright concerns.
Removing the spines allows for faster, cheaper scanning than non-destructive methods, and recent legal rulings suggest that destroying a physical copy after digitizing it may fall under "fair use" because it prevents the original from being resold.
While Amazon states these purchases help improve customer products, booksellers note that the process permanently removes rare historical and intellectual material from circulation, treating specialized knowledge as a mere collection of words for machine learning.