Research AI Minute Newsroom 2026-08-16

AI books sell badly one by one. Together they cut human authors' income in seven genres out of eight.

Researchers classified the full text of 14,419 randomly selected self-published e-books released between January 2023 and March 2026, using the Pangram detector, whose developers report a false-positive rate of 0.04%. Books with substantial AI content make up 20% of that catalogue but only 12.1% of sales and 11.3% of revenue — they perform poorly on their own terms. The market still shifted underneath them: between the first quarter of 2023 and the first quarter of 2026 the catalogue grew 38.3-fold, the number of titles selling each quarter grew 19.2-fold, but quarterly revenue grew only 8.9-fold. Looking only at books with no detected AI text, revenue fell in seven of eight genres even as their average price rose, and AI-heavy titles took a growing share of the scarce top-ranking positions. The effect is strongest in genres with high AI adoption and Kindle Unlimited availability.

Why it mattersThe usual worry about AI writing is that it will get good enough to beat human work. This paper describes a different mechanism, and a harder one to legislate against: individually mediocre books can still damage a market simply by arriving in enough volume to bury everything else in the ranking. The paper is aimed squarely at the fair-use argument that AI output causes no market harm when it does not compete on quality.

✓ Verified · 3 sources

AI books sell badly one by one. Together they cut human authors' income in seven genres out of eight.

Related stories