AI Books: 20% of Amazon Self-Published Catalog, Just 12% of Sales
A new study of Amazon's self-published e-book market finds AI-generated books now comprise 20 percent of the catalog but earn only 12 percent of sales. Analyzing 14,419 titles from January 2023 to March 2026 using a robust detector, researchers show revenue per book has declined even for human-written titles in seven of eight genres, suggesting dilution by volume rather than quality. The exception is Fantasy/Horror, where AI adoption lagged and human titles saw a 35 percent revenue bump. While observational and associational, the findings could provide copyright plaintiffs with concrete market-harm data in cases against AI companies. The study also reveals concentrated AI output among a few prolific producers, with top earners generating millions in gross revenue.
AI Books: 20% of Amazon Self-Published Catalog, Just 12% of Sales
AI-generated books make up 20 percent of Amazon's self-published catalog but bring in only 12 percent of sales. A new study finds that revenue per book is dropping for human-written titles too, in seven of eight genres.
Key takeaway
AI books are flooding Amazon self-publishing through sheer volume, dragging down revenue per book even for human authors, with concentrated production among a few prolific AI-assisted pseudonyms.
What happened
The study analyzed 14,419 randomly selected self-published e-books on Amazon from January 2023 to March 2026, using full-text AI detection with the Pangram v3.3 detector (false-positive rate 0.04 percent). Books with substantial AI content accounted for 20 percent of the catalog but only 12.1 percent of sales and 11.3 percent of revenue, while no-AI books held 62.9 percent of the catalog and generated 72.5 percent of revenue.
Revenue per book dropped for human-written titles in seven of eight genres when comparing 2023 and 2025 releases, with Fantasy/Supernatural/Horror as the only exception, where no-AI titles saw a 35 percent increase. The share of new Top 25 entries with substantial AI content rose from near zero to 31 percent, and a handful of prolific producers (385 identities, 287 increasing output) generated substantial revenue, with the highest-grossing pseudonym earning $1.7 million and the top AI book earning $643,000 on 80,431 copies.
Evidence
AI-generated books make up 20 percent of Amazon's self-published catalog but bring in only 12 percent of sales.
The Decoder · attributed
AI-generated books make up 20 percent of Amazon's self-published catalog but bring in only 12 percent of sales.
Revenue per book is dropping for human-written titles too, in seven of eight genres.
The Decoder · attributed
Revenue per book is dropping for human-written titles too, in seven of eight genres.
The share of new Top 25 entries with substantial AI content rose from near zero to 31 percent over the study period.
The Decoder · attributed
The share of new Top 25 entries with substantial AI content rose from near zero to 31 percent over the study period.
Why it matters
The market-harm data could strengthen copyright lawsuits against AI companies by quantifying how AI-generated books displace human authors' revenue, but the study's observational design limits causal claims.
Limits and uncertainties
The authors stress that their comparisons are observational and associational, not experimental proof of causation.
The method doesn't trace the origin of individual passages or prove that any specific book was copied; it measures aggregate language overlap.
The study relies on a single AI detector (Pangram v3.3) and may not capture all AI-generated content.
Practical implications
Amazon may need to adjust ranking algorithms or review policies to mitigate dilution of human authors' revenue.
Self-published authors should consider differentiation strategies beyond genre, especially in high Kindle Unlimited categories where AI competition is stronger.
AI companies facing copyright litigation might see this study used as evidence of market harm.
What to watch
Whether Amazon changes its recommendation or moderation policies in response to AI book flooding.
Whether courts cite this study in pending AI copyright cases.
Future quarterly reports on revenue per book trends across genres.
