‘They’re likely to get squeezed’: AI slop books are flooding online marketplaces—and it’s coming at the expense of paychecks for human authors

A small book-industry scandal in March turned into a much bigger warning sign. Hachette pulled the horror novel Shy Girl from its U.S. release after accusations that author Mia Ballard had used AI to write it. Ballard denied that, but the book disappeared from Hachette’s site, Amazon, and U.K. shelves anyway. Since then, the fight has shifted from one title to the flood around it.

New research suggests that flood is already substantial. An Originality.AI report found that among more than 2,000 religion books published on Amazon in the past year, 1,272 were flagged as likely written by large language models. A separate Stony Brook University-led study looked at 14,419 self-published genre-fiction books sold on Amazon between 2023 and 2026 and ran them through Pangram, an AI-detection tool. It found 2,880 books with “substantial” AI text, meaning more than 25% of the prose was flagged as generated by AI. More than 2,168 crossed the 50% mark, and nearly 1,000 were more than 90% AI-generated.

The researchers say the effect isn’t just aesthetic. Over those three years, the number of books with observed quarterly sales rose 19.2-fold, while quarterly revenue rose only 8.9-fold. That gap points to lower revenue per book, which is a polite way of saying more titles are chasing the same readers and not enough money is making it back to the people doing the writing. Paramveer Dhillon, a coauthor of the study and an associate professor at the University of Michigan, said the crowding-out effect is likely to squeeze human authors’ revenues.

That pressure is amplified by how Amazon works. The company requires self-published authors to disclose AI use and caps uploads at three titles a day, and a spokesperson said it uses proactive and reactive measures to detect and remove violating content. But Tuhin Chakrabarty, a Stony Brook assistant professor and another study coauthor, said the site’s own recommendation machinery can turn into an “algorithmic bubble,” where AI books get more clicks, more visibility, and then even more clicks. For an unknown human author, that means a book written with actual sweat and tears may never get a fair shot in the first place.

Booksellers are still trying to draw lines. Barnes & Noble’s James Daunt has said the chain will not knowingly stock AI-generated books and wants publishers to label them, but he also made clear he doesn’t want to get lost in arguments over whether a book is 25% AI or 50% AI. That’s probably the real problem here: once the slop starts stacking up, simple rules get messy fast.