An analysis of 14,419 self-published Amazon e-books released between January 2023 and March 2026 puts hard numbers on what many authors have felt for years: the platform is drowning in AI-generated titles, and the flood is dragging down earnings for everyone — including writers who never touched an AI tool.
The researchers, using full-text detection rather than the preview-based methods of earlier studies, sorted books into three bands based on the share flagged as AI-generated by the Pangram v3.3 detector. Books with substantial AI content now make up 20 percent of the catalog studied. They generate just 12.1 percent of sales and 11.3 percent of revenue. Meanwhile, books with no detected AI text represent 62.9 percent of titles and pull in 72.5 percent of revenue. On the surface, that looks like AI books are low-selling filler. The deeper pattern is more troubling. Between Q1 2023 and Q1 2026, the catalog ballooned 38.3x while quarterly revenue grew just 8.9x. More books are fighting over a revenue pool that’s expanding far more slowly.
The most striking finding: revenue per book fell in seven of eight genres even when isolating titles with zero detected AI text. That undercuts the argument that averages are dropping simply because AI slop is padding the catalog. The authors call this “dilution” and are careful to note their comparisons are observational, not proof of causation. The lone exception is Fantasy/Supernatural/Horror, where AI text arrived latest and gained the least traction — revenue per book there actually rose 35 percent for human-authored titles.
The concentration of AI output is stark. Of 385 author identities publishing substantial-AI books, 287 increased their monthly output after their first AI title. The highest-grossing pseudonym pulled in $1.7 million across eight titles. One AI-heavy book sold 80,431 copies for $643,000. And the top AI books overlap heavily with rare language from existing works — 45 percent of their text matched rare expressions found in five or fewer Google Books volumes, compared to 19.1 percent for award-nominated fiction. That pattern echoes earlier reporting on prolific AI-assisted authors like “Coral Hart,” who published over 200 romance titles under 21 pen names in a single year. The economics of volume publishing clearly favor those willing to automate, and the pressure on traditional authors won’t ease unless platforms change how they surface and compensate content.
