Stony Brook study: AI books squeeze authors on Amazon

The alarm inside publishing is not new. In March, Hachette pulled the horror novel Shy Girl from its U.S. release after allegations that its author, Mia Ballard, used AI, even though Ballard denied it. The episode forced booksellers to confront a question they still have not settled: what to do about AI-generated books as they spread across online marketplaces.

The two studies cited here try to put numbers on that question. An Originality.AI report found that of more than 2,000 religion books on Amazon over the last 12 months, 1,272 titles — 63% — were flagged as likely written by large language models. The Stony Brook-led study, which has not yet been peer-reviewed, went further and connected the volume of AI books to money: sales volume grew faster than revenue, which the researchers say left human authors with fewer sales and less income. Tuhin Chakrabarty points to an "algorithmic bubble" in which AI titles get prioritized in search and recommendations, pushing lesser-known human books out of related listings.

Where this settles may depend on choices already being made. Barnes & Noble CEO James Daunt says the retailer does not sell AI books as far as it is aware and demands that publishers label them, but he also says he would not outright ban AI books if they are clearly identified. As AI text improves, the practical test for authors, retailers, and readers may be whether clear labeling is enough to keep human-written books visible — a question the industry has not yet answered.