AI's Double-Edged Sword: Reshaping Book Discoverability

Updated Oct 2026: this editorial has been rewritten and now cites its sources.

Readers increasingly discover books through online stores, recommendation systems and review platforms rather than at a bookshop table. Generative AI has entered that system in two ways. It provides new tools for describing and recommending books, and it has also been used to flood listings and reviews with material that misleads readers. This editorial examines documented cases and the responses to them, and considers what they mean for the discoverability of books by real authors.

Fake books under real names

In August 2023 the Guardian reported that Amazon removed five books that had been listed for sale under the name of publishing writer Jane Friedman. Friedman said she had not written them and believed they had been generated by AI. The books were taken down after she raised the matter publicly. The Guardian reported Friedman's account that she had to fight Amazon's refusal to remove the misattributed titles because she had not trademarked her name.

The case illustrates how AI can damage discoverability for a real author. A reader searching for Friedman's work could have found and bought books she did not write, and the reviews of those books would have reflected on her name.

Platform responses

Amazon responded in two ways that autumn. It required self-publishers to declare AI-generated content when publishing through KDP, and it limited accounts to creating three new titles a day, a step the Guardian connected to concerns about AI-generated books. In announcing the disclosure rule, Amazon said it remained "committed to providing the best possible shopping, reading, and publishing experience for our authors and customers."

A daily limit passes no judgment on any individual book, but it caps how quickly a single account can add listings. A disclosure requirement gives Amazon information about how a book was made.

Fake reviews and the law

Discoverability also depends on reviews, and reviews can be faked with or without AI. The Federal Trade Commission's rule on consumer reviews and testimonials, effective 21 October 2024, prohibits fake reviews, including reviews that misrepresent that they come from a person who does not exist. It also bars buying positive or negative reviews.

The publishing world has already encountered the human version of this problem. In 2023 Del Rey cancelled the debut of Cait Corrain after she admitted using fake Goodreads accounts to manipulate ratings.

Volume makes discovery harder

The pressure on discoverability is rising along with output. Bowker figures reported by Publishers Weekly show that self-published print and ebook titles rose 38.7% in 2025 to more than 3.5 million, and that total US output with ISBNs passed four million. Bowker's Andrew Kovacs attributed the rise in part to the growing availability of new tools that help authors write, publish, distribute and market books. PW also reported that self-published authors appear to have recognized the importance of BISAC subject codes for discovery.

In a market of that size, accurate categories and descriptions are among the few signals a real author controls directly.

Human curation remains a powerful counterweight. NPR reported in 2019 that, according to NPD BookScan, the 26 picks of Reese Witherspoon's book club had sold almost 700% better than the average. Recommendations from a named, accountable person still cut through volume in a way that anonymous listings and ratings do not.

Training data and the author's name

AI also reaches authors' work before any book is listed. In September 2023 the Authors Guild and 17 authors, among them John Grisham, Jodi Picoult and George R.R. Martin, filed a class action against OpenAI on behalf of fiction writers whose works were used to train GPT. The case remains a central test of whether training on published books requires permission.

The other edge: legitimate uses

AI is not solely a source of problems. Authors and publishers can use tools to draft descriptions, test keywords or summarize reader feedback, provided that the results are checked. KDP's own rules distinguish AI-generated content, which must be declared, from AI-assisted content, which need not be. The dividing line is whether the tool produced the content or helped a person produce it. The Authors Guild's AI best practices add that AI-generated text included in a work submitted for copyright registration must be disclosed and disclaimed in the application, and that because book contracts warrant that a manuscript is original to the author, AI-generated text may breach that warranty.

For discoverability, the quality of metadata matters regardless of how it was drafted. KDP's metadata guidelines apply to every listing, and accurate titles, descriptions and categories help readers find the right books.

Protecting a real author's visibility

The documented cases point to several practical steps:

• Monitor listings under the author's name on major retailers and on Goodreads, as the Friedman case demonstrates.
• Keep an official book list on the author's website so that readers can verify titles.
• Report suspicious listings and reviews to the platform.
• Never buy or fake reviews, which the FTC rule prohibits. The Corrain case shows the professional cost as well.
• Follow disclosure rules whenever AI generates content.

A system still adjusting

The cases since 2023 show AI cutting both ways in book discovery. It has been used to impersonate authors and could be used to fake reviews, and platforms and regulators have answered with disclosure rules, listing limits and enforcement powers. For authors, the most reliable protection remains a clear, verified presence that readers can trust.