AI生成书籍涌入亚马逊 人类作者单书收入普遍下滑

This article presents key findings from new research on the flood of AI-generated books on Amazon that are squeezing out human authors. Key takeaways are as follows:

The study analyzed over 14,000 self-published e-books on Amazon listed between January 2023 and March 2026. It found that 20% of sampled books have an AI-generated content share exceeding 25%, yet this group only contributed 11.3% of total revenue. By contrast, fully human-authored books with no AI content account for 62.9% of the sample and generate 72.5% of total revenue. Over the study period, Amazon’s total self-published catalog grew 38.3 times, while total self-publishing revenue only expanded 8.9 times, meaning supply growth far outpaced revenue growth. Human-authored books saw per-title revenue decline in 7 out of 8 studied categories; only the fantasy, supernatural and horror category, where AI penetration has been the slowest, recorded per-title revenue growth.

On an industry level, AI-generated books now account for 31% of the top 25 spots on Amazon’s bestseller lists, up from nearly zero. The turnover rate of bestseller lists has accelerated sharply. One high-output creator has already earned $1.7 million from just 8 AI-generated books. Currently, Amazon requires authors to proactively disclose their use of AI, but does not share this information with consumers. It has also imposed a daily publishing limit of 3 books to crack down on violations such as misappropriation of celebrity authors’ identities.

This article outlines market shifts driven by AI penetration in the content publishing industry, offering valuable insights for brands entering the content space. Key takeaways are as follows:

In terms of market trends, AI content has reached large-scale production and entered mainstream publishing. The industry as a whole is experiencing a "dilution effect" where supply growth far outpaces revenue growth, and per-title revenue for human original content has declined in most categories. Only niche categories with late and low AI penetration still offer room for growth — in this study, for example, human-authored books in the fantasy, supernatural and horror category saw 35% per-title revenue growth. Brands can prioritize these categories for market entry.

On the competitive front, consumers currently cannot access information about AI content share on books sold on the platform. Original content brands can build competitive barriers by positioning themselves as "100% human-created" as a differentiator. At the same time, leading branded accounts have already generated seven-figure dollar revenue from large-scale AI-powered publishing, so brands can also test using AI to cut costs, boost efficiency, and scale content output. Additionally, data shows higher-revenue AI content has higher textual overlap, meaning originality remains the core barrier for long-term brand competition.

This article shares the latest state of Amazon self-publishing and platform policies, outlining clear opportunities and risks for content publishing sellers. Key takeaways are as follows:

For policy and compliance, Amazon currently requires sellers to proactively disclose AI usage when publishing works, but this information is not shared with consumers. The platform has also set a daily publishing limit of 3 books for self-publishing to crack down on violations including bulk AI publishing and misappropriation of well-known authors’ names and writing styles. Sellers must comply with these rules to avoid losses from violations.

In terms of opportunities, AI publishing has already proven profitable. One anonymous account earned $1.7 million from just 8 AI books, with the top-performing single AI book grossing $643,000. Meanwhile, in categories with low AI penetration like fantasy, supernatural and horror, human-authored books still saw 35% per-title revenue growth. Sellers of all positioning can find suitable category opportunities to target.

For risk management, overall industry book supply growth far outpaces total revenue growth. The dilution effect has pushed per-title revenue down in most categories, and bestseller list turnover is accelerating. The share of non-AI books remaining in top rankings has fallen as low as 28%. Sellers should control input risk and conduct thorough product testing before scaling.

The mass proliferation of AI-generated books on Amazon has created new demand shifts and business opportunities for upstream and downstream publishing factories. Key takeaways are as follows:

On the demand side, AI has driven explosive growth in self-publishing. Over the study period, Amazon’s total self-published catalog grew 38.3 times, and the number of active books per quarter grew 19.2 times. AI books are characterized by bulk production, small-batch multiple runs, which is completely different from the large print runs of traditional publishing demand. This puts forward new requirements for factories’ flexible production capabilities.

In terms of business opportunities, AI book output is concentrated among a small number of high-output accounts: one creator published over 200 romance novels under 21 different pen names in a single year. These individual creators have strong demand for small-batch, fast printing and binding services. Factories can develop targeted small-batch fast-print service packages to serve the large pool of individual creators and capture the growth dividend of self-publishing.

For long-term development, AI has already pushed the publishing industry toward digitalized, large-scale production. Factories need to accelerate digital transformation for order matching and production management to adapt to the new demand for small-batch fast turnaround. This is the only way to access the new publishing ecosystem and unlock more growth opportunities.

This article outlines the latest development pain points and trends in the AI publishing industry, pointing out clear business directions for publishing-related service providers. Key takeaways are as follows:

In terms of industry development trends, AI-generated content has achieved large-scale penetration in self-publishing, with AI books now accounting for 31% of new entries to top bestseller lists. The clear trend of large-scale production has spawned massive new service demand around AI publishing, creating broad market space.

For core customer pain points: First, platforms, human authors and copyright holders all need accurate AI-generated content detection. While low false-positive detectors like Pangram v3.3 already exist, AI content lookup services for general consumers remain a gap in the market. Second, AI copyright litigation has long lacked empirical evidence of market dilution caused by AI content, and this study fills that gap, providing new foundational data for legal services and copyright authentication. Third, creators have demand for improving AI content production efficiency: leading high-output creators have already dramatically increased monthly output with AI, driving strong demand for related tooling services.

Service providers can develop targeted solutions around these pain points to capture early market share.

This article organizes the challenges self-publishing platforms face following the influx of AI-generated books, summarizes existing platforms’ response experience, and provides a reference for platform operation and management. Key takeaways are as follows:

Core challenges brought by AI books for platforms today: Bulk AI publishing has caused explosive growth in platform book supply, far outpacing total revenue growth. This dilution effect has led to widespread per-title revenue declines for human authors, squeezing their livelihoods. It has also triggered problems including AI books misappropriating famous authors’ names and writing styles, infringing copyright, leading to frequent copyright disputes. Prior to this study, there was a lack of empirical evidence quantifying AI’s market impact in such cases.

Mature response experience from existing platforms: Amazon currently requires authors to proactively declare AI usage when publishing works, and has set a daily self-publishing limit of 3 books to curb bulk non-compliant publishing. This response framework can serve as a reference for similar content platforms.

Areas for improvement: Platforms currently do not disclose AI content share information to consumers, failing to meet consumers’ right to know. Going forward, platforms can improve their information disclosure mechanisms to balance the interests of AI creators and human original authors, and maintain long-term stable platform ecosystem health.

This article shares new quantitative research findings on AI publishing on Amazon, raises new industry questions, and provides rich empirical data for researchers in publishing industry and AI copyright research. Key takeaways are as follows:

New industry trends and new business models: AI-generated content has achieved large-scale commercial entry into the publishing industry. The share of AI books among new entries to top bestseller lists has risen from nearly zero to 31%. AI book output is concentrated among a small number of high-output accounts, forming a new business model of "bulk AI content generation + self-publishing platform listing for profit". Multiple accounts have already generated seven-figure dollar revenue from this model, which has already been validated in the AI music industry and has strong replicability, making it worthy of further in-depth research.

New research questions and contributions: This study puts forward the concept of the "dilution effect" of AI content. Through controlled research, it proves that after the large-scale influx of AI content, even fully human original content sees widespread per-title revenue declines. This finding fills the empirical gap on the market dilution impact of AI content in copyright litigation, making it a valuable reference for AI copyright compliance and policy research.

Additionally, the study identified a new linguistic pattern in AI content: for AI content, every 10-fold increase in revenue corresponds to a 7.6 percentage point increase in textual overlap of rare expressions. No such correlation exists in human original content. This provides a new direction for AI content detection research.