A new academic study tracking 14,419 self-published books on Amazon found the platform's catalog grew 38.3 times larger between 2023 and 2026, while quarterly revenue across those same books grew only 8.9 times. Human authors, including those who never used AI at all, are seeing revenue per book fall as a result, according to the researchers.
The study, titled "Generative AI floods and dilutes the market for books," comes from researchers at Stony Brook University, Columbia Law School and the University of Michigan. They matched 14,419 self-published genre-fiction ebooks to daily sales records running from the first quarter of 2023 through June 2026, according to Startup Fortune. The paper has not yet been peer-reviewed, Fortune reported.
The researchers ran the books through AI-detection software called Pangram. Books with "substantial" AI content, meaning more than a quarter of the text was flagged as machine-generated, made up 20% of the catalog, or 2,880 titles, Fortune reported. Of those, 2,168 had more than half their prose flagged as AI-generated, and nearly 1,000 novels were flagged as more than 90% AI-generated.
Despite that volume, AI-heavy books earned a smaller slice of the pie: 12.1% of sales and 11.3% of revenue, according to Startup Fortune. Books with no detected AI text still made up the majority of the catalog, 62.9%, and pulled in 72.5% of revenue.
Selling titles per quarter climbed 19.2 times over the period studied, Fortune reported, a pace far outstripping revenue growth. The researchers describe the relationship between AI volume and shrinking per-book revenue as associational rather than proven causation, according to Startup Fortune.
The effect landed on human authors broadly, not just those competing directly with AI-generated titles. In six of eight genres studied, revenue per book fell when comparing titles released in 2023 against titles released in 2025 over the same post-release window, according to Startup Fortune. Narrowed to books with no AI text at all, revenue fell in seven of eight genres.
Why volume beats quality on Amazon's rankings
Paramveer Dhillon, an associate professor of information at the University of Michigan and one of the study's coauthors, told Fortune that the flood of AI books crowds out human authors even when much of that content is low quality.
"The crowding-out effect generated by these books—most of which are low quality, but some of which are still good enough that they lead to higher ranks—due to this increased competition, they're likely to get squeezed, and their revenues are reduced," Dhillon told Fortune.
Tuhin Chakrabarty, an assistant professor of computer science at Stony Brook University and another coauthor, told Fortune that the "algorithmic bubble" of online marketplaces bears some of the blame. He told Fortune that a book written by an author who is not well known can fail to appear in a site's related-books listings because so many AI-generated titles are surfacing against the same keywords and search terms.
Separate coverage from outlets including The Decoder and the Blog Herald, cited by Startup Fortune, pointed to another finding from the same dataset: AI-written fiction now fills roughly 31% of new slots on Amazon's bestseller lists in the genres the study tracked.
A separate report from Originality.AI, published last month and cited by Fortune, found that of more than 2,000 religion books published on Amazon over the prior 12 months, 1,272 titles, or 63%, were flagged as likely written by large language models.
Retailers and rules already in place
Amazon requires self-published authors to disclose AI-generated content and limits uploads to three new titles per day per author, according to both Fortune and Startup Fortune. Startup Fortune reported that the upload cap was introduced in 2023. Neither measure has slowed the catalog's growth, Startup Fortune noted.
An Amazon spokesperson told Fortune the company is "committed to a bookstore that readers love and publishers and authors trust, with multiple safeguards we continuously strengthen to protect their experience." The spokesperson said Amazon's content guidelines govern which books can be listed and that the company takes "proactive and reactive measures to prevent, detect, and remove content that violates those guidelines, whether AI-generated or not."
Barnes & Noble has taken a harder line. CEO James Daunt told Fortune in May that the retailer does not sell AI-generated books and excludes them from its catalogue. "This is a straightforward rejection of AI books," Daunt told Fortune, though he acknowledged that drawing a line at, say, 25% or 50% AI-generated content would be difficult to enforce in practice.
The dispute over AI and books has already reached the courts. In July, a federal judge approved a $1.5 billion copyright settlement requiring Anthropic to pay thousands of authors after the company downloaded and stored millions of pirated books to train its Claude chatbot, Fortune reported. Authors have also sued OpenAI, Meta and xAI over similar allegations.
Chakrabarty told Fortune he expects the market to split rather than collapse: a segment of formulaic, AI-generated work that proves highly profitable, alongside renewed demand for literary fiction that reflects writing AI cannot yet match. "So I guess human authors will find which clan they want to belong to," he told Fortune.
Sources
Every quotation and figure in this article is checked against the source reporting linked above before publication. How we report