AI-Generated Books Reduce Human Authors' Market Share
Business Insider reported on an analysis of more than 14,000 Amazon books sold between 2023 and 2026, finding that books containing substantial AI-written text expanded faster than spending on them. The number of such books with sales rose 19-fold while spending rose ninefold, and books without detected AI text lost market share, adding evidence to debate over whether generative AI can harm markets for human-authored works.
AI-generated books are expanding on Amazon faster than spending on them, reducing the market share held by titles without detected AI text, according to a recent analysis cited by Business Insider. The study examined more than 14,000 books sold on Amazon between 2023 and 2026.
Business Insider reports that the number of books with sales that contained substantial AI-written text rose 19-fold during the period, while reader spending on those books increased ninefold. The gap indicates a much more crowded market without commensurate growth in total spending, leaving less revenue per title on average. The researchers also found that books without detected AI text lost market share as AI-generated books spread.
Market harm enters the copyright debate
The findings bear on a central question in US generative AI copyright disputes: whether AI systems and their outputs damage markets for copyrighted works. Business Insider notes that the US Copyright Office previously identified market substitution as a weakness in broad fair-use arguments for commercial generative AI, particularly when generated content competes with the source material.
Business Insider also reports that Anthropic bought millions of used books, removed their spines, and scanned them for model training. A federal judge ruled that digitizing lawfully acquired books in that process was fair use, according to the outlet. That ruling concerned the copying and digitization process, while the newer research focuses on the downstream market in which AI-generated titles compete for readers and sales.
Public reporting has documented the broader demand for pre-2022 books as model-training material. The Atlantic reported that court documents unsealed in litigation showed Anthropic bought millions of new and used books in spring 2024 and destroyed them for scanning. The article noted that books published before ChatGPT's 2022 release are attractive because they are less likely to contain AI-generated text.
Data quality and physical-book sourcing
404 Media reported in July that ISBNdb, a book database and sourcing company, marketed high-volume print-book acquisition for AI training. The company described books as curated, peer-reviewed, domain-specific human knowledge in material quoted by 404 Media. Mashable reported that ISBNdb had also advertised destructive scanning and acknowledged that the optics problem was real.
Key Points
- 1Business Insider's cited study links rapid growth in AI-written Amazon books to declining market share for titles without detected AI text.
- 2The evidence shifts debate from training-data copying alone toward measurable downstream substitution in commercial content markets.
- 3The reporting highlights the value of edited, human-authored books as training data and the controversy surrounding their acquisition and destructive scanning.
Scoring Rationale
The reported market-share evidence is directly relevant to copyright litigation and policy arguments over generative AI's downstream economic effects. It also highlights provenance and content-supply challenges for platforms and teams deploying text-generation systems, though the evidence is limited to a book-market sample.
Sources
Public references used for this report.
Practice with real Retail & eCommerce data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Retail & eCommerce problems
