Why AI Companies Are Buying—And Then Destroying—Old Books

AI developers are acquiring vast collections of physical books, only to digitize and then destroy them. This practice aims to create unique, high-quality training datasets for AI models, distinct from the often-unreliable text found online.
Key takeaways
- AI firms are using physical books for training data.
- Digitizing and destroying books creates specialized datasets.
- This aims to improve AI model quality and reliability.
- Expect more accurate AI outputs from these models.
Why it matters
This trend means AI models may soon be trained on more curated, human-authored content, potentially leading to more accurate and reliable AI assistant outputs. Users can expect improved performance in tasks requiring factual accuracy and nuanced understanding.
Try this on SynaBot
Related AI assistants, prompts, and tools from the SynaBot catalog.
- LongShotLongShot helps individuals and teams generate factual, SEO-optimized content for copywriting, marketing campaigns, ads, and sales enablement, accelerating content creation and ideation.
- NextThreeBooksNextThreeBooks — Discover your next favorite book with AI-powered, personalized suggestions. It sits in the hr & recruiting category and is built to assist with sourcing, screening, interview prep, onboarding content, and internal HR processes.
- LongShot AILongShot AI specializes in generating long-form, factual content verified for accuracy. It helps create detailed blog posts, articles, and research papers, ensuring quality and credibility.



