AI Books Are Taking Sales From Human Authors
Studies show books built with generative AI now flood Amazon, with about one in three self published genre novels carrying AI text and human authors losing revenue.

A flood of books written with AI
Books written with generative AI are reaching the market in such abundance that studies now show them hurting the revenue of human authors. Morning Brew reports that of the roughly 2,000 religion books published to Amazon in the last year, 63 percent were flagged as likely written by large language models, according to a study by Originality.AI. That is not a few hobbyists experimenting. That is most of a category.
A second report, led by Stony Brook University, looked at 14,419 self published genre fiction books sold on Amazon between 2023 and 2026. More than one third had at least 20 percent AI generated text, and about 1,000 of those novels were likely almost entirely AI generated. The books arrive fast, they arrive cheap, and they arrive whether or not anyone is waiting to read them.
The reader pool did not grow with the book pool
The Stony Brook study concluded that AI books have reshaped the market, lowering the overall revenue available to human authors. The logic is simple and a little brutal. AI titles are produced so quickly and so cheaply that the pool of readers cannot keep up, so every new flood of titles divides the same attention and the same dollars into thinner slices. The human author who spent a year on one novel now competes with a catalog built in a weekend.
Amazon told The New York Times that it regulates AI books, although it does not prohibit their sale, and it allows authors to publish up to ten AI books a week. Ten a week. At that pace, the contest is no longer only about who writes the best book. It is about who can occupy the most shelf space before a reader ever makes a choice, and shelf space on a screen is just search results and recommendation rows.
Every creator market runs on the same math
If you make anything for a living, this story is also about you. The same volume math now presses on music, video, photography, and design. A client can generate a hundred serviceable images before lunch. A channel can publish a video every day without filming anything. None of that output needs to be great to do damage. It only needs to be present, priced near zero, and good enough to make a buyer hesitate before paying a person.
What volume cannot copy is a voice with a history. Readers follow authors whose judgment they trust. Clients hire creators whose eye they have watched work in public, at real events, with real people whose names and faces are attached. A model can imitate a style. It cannot have stood in your city, shot your festival, or built a reputation one delivered job at a time. That lived record is the asset, and it compounds every time you publish work only you could have made.
Use the tool and keep the authorship
The practical answer is not to refuse the tool. It is to put the tool in its place. Let AI handle outlines, research summaries, first pass edits, and the admin work around the work. Keep the choices that carry taste, memory, and accountability in human hands, and say so plainly to your audience. Buyers who care about authorship can only choose it if they can see it, so show the process, the drafts, the locations, and the people.
Then build the part of the business a flood cannot wash out. Sell the community around the work, the access, the live experience, and the trust, not only the finished file. A book, a video, or a photo package can be copied in spirit by a machine in an afternoon. A room full of people who believe in the person who made it cannot. Authors learned that this week on Amazon. Creators everywhere else should take the note early.
Sources and further reading
Sources were checked for this article. Tool features, prices and availability can change. Product claims are attributed to their publishers.


