Can you make money publishing AI books on Amazon?

After AI jingles for brands, we return to a catalog-based revenue model: publishing books written (fully or partly) with AI on Amazon Kindle Direct Publishing (KDP). It's a real business with solid general data on what an average indie author earns — but also the only one in this entire cluster where the risk isn't legal or platform-related, but a real physical safety risk to the reader.

The best-documented case: Coral Hart

The New York Times documented the case of Coral Hart, an author who published over 200 romance novels in a year under 21 different pen names, collectively selling around 50,000 copies and, per the report, earning six-figure revenue. It's worth stating with nuance: other writers and independent analysts publicly questioned that "six figures" claim, pointing out the math isn't entirely clear from the available data. It's worth treating the case as the best-documented in the category, not as a 100% verified figure.

What an indie author actually earns

The Alliance of Independent Authors, in its 2025 income survey, reported a median of USD 13,500 a year for self-published authors (not AI-specific), against USD 6,000-8,000 for traditionally published authors. Industry sources (specialized blogs, not independently audited) add more detail by catalog size: with 3-6 books and systematic advertising, USD 500-2,000/month is a realistic 4-8 month target; full-time income (USD 5,000-10,000+/month) typically requires 6-12 books and 6-12 months of consistent execution; authors with 25+ books report a median of ~USD 3,000/month, while around 80% of authors with 1-3 books earn under USD 100/month. Catalog size, not writing quality or genre, is the variable that most separates income tiers.

What Amazon's AI policy actually requires

Amazon requires disclosure, via a private checkbox in the publishing form, if the book contains AI-generated text, images or translations — that disclosure never appears on the product's public page. The key distinction: if AI created the content, you must disclose it; if AI only helped refine content the author wrote (brainstorming, editing, error-checking), no disclosure is needed. Amazon also caps uploads at 10 per week per book format, specifically targeting mass-produced, chain-generated low-content books — publishers who systematically upload undisclosed AI content, or low-quality books, face account-wide publishing restrictions.

The risk no other article in this series had: real physical safety

Here the risk stops being legal or platform-related and becomes directly about reader safety. Since 2023, outlets like Civil Eats, CBC, Rolling Stone and Fortune have documented a wave of mushroom foraging guides on Amazon that appear to be AI-generated, with potentially incorrect information about which species are edible — an error in that kind of guide can kill the person following it, not just earn a bad review. In a related case, journalist Talia Lavin discovered AI-generated knockoffs mimicking the title of her own published book, sold on Amazon without her consent. The lesson for anyone considering this business: in niches where incorrect content has real consequences (health, safety, medicine, species identification), publishing with AI without expert human verification is a risk that goes far beyond selling a few copies.

So, is it worth it?

It's a real catalog business, where catalog size — not virality — determines income, with solid general data from the self-publishing industry. It's worth treating as a volume-and-patience business (much like AI music on Spotify), disclosing AI use where it applies to avoid account restrictions, and avoiding entirely any niche where an incorrect guide could put a reader's safety at real risk.