AI Books Are Quietly Killing Author Pay

AI-generated books make up 20 percent of Amazon's self-published catalog but bring in just 12 percent of revenue. The 20-percent figure sounds like the headline, but it's not the real story. The real story is that revenue per book has dropped even for human-written titles with zero AI text , in seven of eight genres. The AI books are not just sitting at the bottom of the market. They are pulling the ceiling down on everyone.

The 38x catalog explosion that no one noticed

Between Q1 2023 and Q1 2026, the cumulative catalog of self-published e-books grew 38.3 times. The number of titles selling each quarter grew 19.2 times. Quarterly revenue grew just 8.9 times. The math is brutal: far more books are competing for a revenue pool that is expanding much slower than the shelf space.

Researchers at the University of Chicago and the Allen Institute for AI analyzed 14,419 randomly selected self-published e-books released in that window. They pulled daily sales figures from an internal dataset maintained by one of the five largest US publishers , a dataset that covers roughly 95 percent of all e-books sold daily on Amazon.

Unlike earlier studies that tried to detect AI text from short book previews, this one classified each title based on its full text using Pangram v3.3, a detector with a 0.04 percent false-positive rate. Books were sorted into three bands: no detected AI text, light (up to 25 percent), and substantial (over 25 percent).

The dilution effect hits everyone

In six of eight genres, revenue per book dropped when comparing titles released in 2023 and 2025 over the same post-release window. Looking only at books with no detected AI text, revenue fell in seven of eight genres.

That last finding is the one that matters. It rules out the easy explanation that the average is dropping because poorly selling AI books are padding the catalog. The authors call this "dilution" , AI-generated volume spreads legitimate authors' earnings across a smaller piece of a crowded pie. They stress these are observational comparisons, not experimental proof of causation. But the pattern is consistent.

The only exception is Fantasy/Supernatural/Horror, where AI text arrived latest and gained the least traction. There, revenue per book for titles with no detected AI text actually rose 35 percent. The researchers say this reversal argues against a general market downturn as the cause.

AI books are breaking into the top ranks

The story gets worse for human authors who are selling well. The share of new Top 25 entrants with substantial AI content rose from near zero to 31 percent over the study period. The top ranks turned over faster. The share of books with no detected AI text that stayed in the Top 25 from one quarter to the next dropped as low as 28 percent at one point, before settling around 62 percent.

The most successful AI-generated authors are not scrappy amateurs. Of 385 author identities that published more titles with substantial AI content after their first AI book, 287 increased their monthly output. The highest-grossing pseudonym earned $1.7 million in gross revenue before platform fees across eight titles. A single highest-grossing book with substantial AI content brought in $643,000 on 80,431 copies sold.

The copyright lawsuit connection

The study feeds directly into ongoing copyright battles. In Kadrey v. Meta, Judge Vince Chhabria ruled in Meta's favor in June 2025 but issued a pointed warning. He said it was hard to imagine that using copyrighted books to build a product generating billions in revenue while producing a potentially endless flood of competing works would qualify as fair use.

What the plaintiffs lacked was empirical evidence that this flood was causing concrete market harm. The current study delivers exactly that evidence: books with no detected AI text earn less as AI titles enter the market in large numbers.

Making the legal case harder for Amazon is the fact that AI content cannot be identified on the platform itself. Authors must disclose AI involvement when publishing through Kindle Direct Publishing, but Amazon does not pass that information along to customers. The platform's main response to date has been capping publications at three per day.

The language overlap question

The researchers also examined how much top-selling AI books draw on rare language from existing works. They used the Allen Institute's infini-gram tool alongside the Google Books index to identify rare expressions that appear in five or fewer Google Books volumes and are entirely absent from a 4.7-trillion-token web snapshot.

Among the 50 highest-grossing titles with substantial AI content, these rare expressions covered 45 percent of the text, compared to 37.7 percent for the top 50 books with no detected AI text. For award-winning fiction, the figure was just 19.1 percent. Within AI books, overlap rose 7.6 percentage points for every tenfold increase in revenue. The method does not trace the origin of individual passages, but the pattern strongly suggests that successful AI books are heavily informed by specific existing works.

The bottom line

The study does not prove that AI-generated books are the sole cause of falling author revenue. Correlation is not causation, and the researchers are careful not to overclaim. But the timing is damning: as AI content surged into the catalog between 2023 and 2026, human authors saw their per-book earnings drop in nearly every genre. When the only category that escaped the trend is also the one where AI content arrived latest, the circumstantial case is strong.

For the copyright lawsuits, this study provides exactly what was missing , numbers. Judges and juries respond to market data, not to anecdotes about unfair competition. Whether that data is enough to shift the legal landscape is a question for the courts. But the argument just got a lot harder for the AI companies to dismiss.