AI Fiction Grabs Top Amazon Ranks, Revenue Per Book Falls

Researchers scanned 14,419 self-published genre-fiction titles sold on Amazon between 2023 and 2026, ran full-text AI detection on each one, and matched the results to daily sales through June 2026, per the arXiv paper. Books where AI wrote more than 25% of the text reached real commercial scale, taking top-rank spots once held by human authors, and none of them disclosed it.

The number of books with observed quarterly sales grew 19.2x over the study period. Quarterly revenue grew only 8.9x. That gap means revenue per selling book fell across most genres, and the drop hit hardest where AI diffusion is heaviest and where Kindle Unlimited, the subscription plan that pays authors per page read instead of per sale, dominates.

The authors call this dilution through scale, not a quality collapse. More books are competing for a reader-spending pool that isn't growing nearly as fast, and undisclosed AI-heavy titles are taking a growing share of the ranks that used to go to humans.

Built for a courtroom

That framing isn't incidental. The paper ties its findings to the market-effect prong of the fair use defense, the legal test asking whether a use harms the market for the original work. Measurable per-book revenue decline in AI-saturated genres is exactly the evidence a plaintiff's expert would want in a suit over AI training data or text produced by machine models.

What the paper doesn't establish is causation between any single AI tool and any one author's lost sales, only a correlation at genre scale. For Kindle Unlimited authors already watching per-page payout rates slide, it's one more number matching their royalty statements.