Study Claims Over Half of Amazon's 2025 Sample Books Contain AI-Generated Low-Quality Text
A sampling study based on English data shows that in 2025, over half of the pages in books sampled from the Amazon platform contain low-quality "slop" text content generated in bulk by generative AI.
The research team classified a large number of mechanized paragraphs, semantically hollow rewrites, and templated chapters as suspected AI-generated content through technical detection of the text structure and repetition patterns in the sampled books, estimating that over 50% of the books in the sample exhibit this type of "information filling" phenomenon.
In terms of market mechanisms, a large volume of low-cost AI-generated text is concentratedly released by small and medium publishers and individual uploaders, quickly occupying search results and recommendation slots with extremely low creative costs, squeezing the exposure and conversion space for real authors, forcing readers to spend more time filtering for high-quality content, while platforms gain a larger "inventory scale" and transaction commissions in the short term.
Source: Public Information
ABAB AI Insight
Historically, the Amazon book platform has been repeatedly named by English media and author organizations for the emergence of phenomena involving the bulk generation of "pseudo-books," "copy books," and "summary books" using AI, including structural plagiarism, rewrite copying of bestsellers, and unauthorized "manualization," which has led the author community to pressure the platform multiple times to label and clean up related content.
In terms of capital pathways, the core motivation for AI-generated "slop" text lies in extreme cost compression on the publishing side: uploaders generate large volumes of books at almost zero marginal cost, occupying long-tail search and recommendation traffic, using small sales or mistaken purchase revenue as cash flow, while the platform continues to extract commissions from each transaction without significantly increasing editorial review costs, achieving a profit structure of "content quantity expansion and asymmetrical review investment."
In analogy and industry positioning, the current issue of AI pseudo-books on Amazon is highly similar to past problems of counterfeit goods, fake orders, and imitation brands on e-commerce platforms: in the early stages, the platform's core narrative was "openness + scale," tolerating or lightly governing inferior content to expand categories and GMV; it was only after the accumulation of rights protection incidents and regulatory pressure that it entered a stronger governance phase, a trajectory that can be found in the governance history of "clickbait content" on online video platforms and "junk apps" in app stores.
In structural judgment, this wave of AI "slop" text diffusion essentially belongs to the overlap of "pricing power transfer + industrial chain reconstruction": when content production costs are driven down to extremely low levels by AI, platforms shift from "curated content distribution" to "mass content aggregation," real creators lose pricing power over the scarcity of their works, and pricing power shifts to platforms that control distribution and search ranking; at the same time, traditional publishing processes such as editing, topic selection, and proofreading are compressed, with some functions replaced by algorithms and automated generation, pushing the book industry chain from "publishing-led" to "platform and algorithm-led" reconstruction.
ABAB News · Cognitive Laws
- When content costs approach zero, pricing power naturally shifts to the distribution end.
- Cheap text fills the shelves, while scarce attention reshapes the value curve.
- The moment creators lose scarcity, the platform takes over the fate of the works.