AI-generated books have flooded Amazon, leaving human authors struggling to reach readers as nearly one in five self-published titles relies heavily on artificial intelligence.
Research published this week analyzed 14,419 self-published e-books released on Amazon between January 2023 and March 2026 to evaluate how automated writing tools are affecting the publishing marketplace.
The study categorized titles by estimated software involvement and found that books containing at least 25 percent artificial intelligence content accounted for roughly 20 percent of all examined e-books.
Amazon has dominated digital reading since launching its Kindle e-reader line, offering authors direct access to readers through its Kindle Direct Publishing platform without needing traditional publishers.
While self-publishing services originally aimed to help independent writers bypass legacy gatekeepers, the open platform model has inadvertently enabled automated accounts to upload low-quality works on a massive scale.
Book releases outpace sales revenue
To assess the financial impact on writers, researchers analyzed an internal sales dataset from one of the largest publishing houses in the United States, tracking about 500,000 Amazon titles.
The dataset covers approximately 95 percent of e-books sold on the site, providing detailed tracking of daily unit sales and overall financial returns across multiple genres.
The data revealed a stark mismatch between production volume and market growth during the research period, as the number of e-books sold per quarter surged 19.2-fold while total quarterly revenue increased by only 8.9-fold.
Because the volume of available titles expanded far faster than total consumer spending, average revenue per book declined across numerous reading categories.
Traditional authors invest substantial time, research, and edit cycles into individual manuscripts, leaving human writers vulnerable when market flooding lowers earnings per title.
Conversely, automated operators generate hundreds or thousands of books using large language models, relying on sheer volume to turn a profit even if individual titles sell only a handful of copies.
Detection flaws complicate industry action
The influx of automated text on self-publishing portals reflects a broader dilemma facing commercial publishers and literary agencies seeking to balance innovation with authenticity.
Industry professionals are exploring generative software for administrative and editorial assistance, but struggle to verify whether incoming manuscripts are truly written by human authors.
Verification is further complicated by technical limitations in automated detection tools, which frequently produce false positive flags when analyzing complex human prose.
Because algorithmic detection cannot serve as definitive legal or editorial proof, publishing organizations and literary groups are actively debating new standards to safeguard readers and authors from artificial intelligence flooding.
