Fact check: AI companies are purchasing large quantities of used and rare books.
Verdict: mostly true — Trust Score 79/100
The post accurately reports on a 404 Media investigation that details AI companies purchasing, scanning, and destroying physical books to train their models. This practice has been confirmed by 4 sources including Tom's Hardware, TNW, and Futurism. The claims about increased sales for booksellers and the targeting of books by ISBN are also corroborated by these reports.

- Platform
- threads
- Source author
- major.leeg — see all fact-checks of this account
- Original post
- https://www.threads.com/@major.leeg/post/DbWYAXsjwHQ?xmt=AQG0fL6_OwS28eTg9H78fsUtVVSU_-ZREk4xhsvlEOpXuYPBKIqfr1CHU0l8YQIhCwWKTqY_
- Verified on
- July 28, 2026
- Verification ID
- Z8BndA1NSW-o6lhkyD0ZGQ
Original content reviewed
Platform: THREADS Author: @major.leeg --- Caption/Description --- BREAKING: AI companies are quietly purchasing large quantities of used and rare books, scanning their contents, and destroying the physical copies, according to a 404 Media investigation. One bookseller told the outlet his weekly sales jumped from about 20 books to hundreds starting in April, with buyers appearing to target books by ISBN rather than content or condition. ---VERIFICATION_SUMMARY--- Platform: THREADS Author: @major.leeg What the video shows (from video analysis): The video shows a close-up of various books being fed into a machine that appears to be cutting or shredding them. The books are of different colors and sizes, and the machine systematically processes each one, destroying its binding and pages. --- HOW TO VERIFY THIS (provenance-first) --- 1. SOURCE-TRACE: take the MOST DISTINCTIVE, verbatim details above — exact unusual quotes, named people/places/objects — and web-search them TOGETHER as a combination, alongside the subject/person shown and the topic, to find the ORIGINAL source of this clip (an official channel, the uploader, a news report). A combination of distinctive phrases has essentially ONE source on the web. 2. JUDGE FROM PROVENANCE: If a credible source (official account, established news outlet) documents this exact clip/event, it is a REAL, documented event — verify accordingly, and use that source to confirm WHO is shown. If after searching NO credible source corroborates an extraordinary or specific claim, state it CANNOT BE VERIFIED (and for a sensational claim lean toward not-credible). Do NOT default to "verified" on tangential, generic facts. 3. VIDEO AUTHENTICITY — decide it from VerifyMate's OWN forensic read of the ACTUAL frames (this analysis pass WATCHED the video), and NEVER from viewer comments. The read: "no strong signs of AI generation (confidence 0%)". Then SOURCE-TRACE the distinctive details (steps 1-2) to confirm whether this is a real, docu
Claims analyzed (6)
- mostly true: AI companies are purchasing large quantities of used and rare books.
Multiple news outlets, citing 404 Media, report that AI companies are indeed buying large quantities of used and rare books. This is driven by the need for high-quality, human-authored training data, especially books published before 2022 to avoid AI-generated 'slop'. - mostly true: AI companies are scanning the contents of these books.
Reports confirm that AI companies are scanning the contents of purchased books to digitize them for training large language models. This process often involves 'destructive scanning' where bindings are removed. - mostly true: AI companies are destroying the physical copies of the books after scanning them.
Multiple sources confirm that after scanning, AI companies often destroy or recycle the physical copies of the books. This practice is sometimes referred to as 'destructive scanning' and has raised ethical concerns. - mostly true: A 404 Media investigation reported these activities by AI companies.
Multiple news articles explicitly cite a 404 Media investigation as the source for the claims about AI companies purchasing, scanning, and destroying books. - mostly true: One bookseller reported his weekly sales increased from approximately 20 books to hundreds starting in April.
Several news outlets, citing the 404 Media investigation, report that a bookseller observed a significant increase in weekly sales, from around 20 books to hundreds, starting in April. This surge is attributed to purchases by AI companies. - mostly true: Buyers of these books appear to target them by ISBN rather than content or condition.
Reports indicate that AI companies, often through intermediaries like ISBNdb, are targeting books by their ISBNs, with little regard for the content or physical condition. This suggests a focus on acquiring unique textual data for AI training.
Sources consulted (6)
- thecanary.co — thecanary.co
- AI companies are buying up old books to escape their own AI slop — vertexaisearch.cloud.google.com
- AI companies are reportedly shredding millions of books after using them to train AI models — tech giants outsource to middlemen to secretly buy up books for training material | Tom's Hardware — vertexaisearch.cloud.google.com
- visiontimes.com — visiontimes.com
- aa.com.tr — aa.com.tr
- thenextweb.com — thenextweb.com
Related verifications
- Elon Musk stated that our AI provider CEO Sam Altman is an untrustworthy individual. — Verified 90/100threads · view post ↗
The post accurately reports that Elon Musk stated our AI provider CEO Sam Altman is untrustworthy and accused him of "stealing an open source nonprofit." These statements are confirmed by 4 sources in…
- A Trump-appointed federal judge just REJECTED Trump's DOJ attempt to erase Steve Bannon's… — Mostly true 65/100threads · view post ↗
The post accurately reports that a Trump-appointed federal judge, Carl Nichols, rejected the Department of Justice's attempt to erase Steve Bannon's contempt-of-Congress conviction. The DOJ had argued…
- Nvidia is in talks to acquire Hugging Face for $13,000,000,000. — Verified 85/100threads · view post ↗
The post accurately states that Nvidia is in talks to acquire Hugging Face for over $13 billion, a detail confirmed by 14 sources. Hugging Face is indeed an AI platform, and an our AI provider agent p…
- A male individual (Milo Yiannopoulos) was detained by ICE (U.S. Immigration and Customs E… — Mixed 45/100threads · view post ↗
The X.com post attributed to Milo Yiannopoulos, stating "Total legal immunity for ICE agents when they're on duty," is authentic and was posted on September 26, 2025. However, the Threads caption clai…
- Cheesecake Factory ($CAKE) is outperforming Nvidia ($NVDA) by 109% year to date. — False 25/100threads · view post ↗
The post falsely claims that The Cheesecake Factory is outperforming Nvidia by 109% year-to-date. While The Cheesecake Factory has shown a strong year-to-date return of approximately 54.40% as of June…
AI-generated analysis. Not a substitute for professional fact-checking.