AI-generated books now account for 20% of Amazon’s self-published catalog but only 12% of sales, according to a study reported by The Decoder. The same research found that revenue per book is falling for human-written titles in seven of eight genres.
The finding matters because it moves the AI publishing debate from quality complaints to market effects. A flood of cheap or quickly produced titles can make discovery harder even when those books do not sell especially well.
For human authors, the risk is not only direct substitution. If readers and recommendation systems must sort through more low-performing inventory, legitimate books may receive less visibility and lower revenue.
The study could also matter in copyright lawsuits, where plaintiffs often need evidence of market harm. The numbers do not prove every AI book is infringing or low quality, but they show that synthetic publishing can change the economics of self-publishing at scale.