Researchers Find Bestselling Novel Daggermouth May Contain Substantial AI-Generated Text
Key Takeaways
- ▸AI-generated fiction may be achieving mainstream success: Daggermouth became a bestseller despite potential AI involvement, suggesting AI writing quality has improved significantly
- ▸AI detection tools have limitations: While Pangram flagged Daggermouth at 60% AI-written, researchers acknowledge the tool cannot definitively prove AI use, and authors can credibly deny involvement
- ▸Amazon publishing faces a quality crisis: New releases nearly tripled from 2022-2025, with research indicating one in five Kindle ebooks contains 25% or more AI-generated text
Summary
A new study by Stony Brook University researchers examined over 14,000 self-published Kindle ebooks using Pangram, an AI detection tool, and discovered that Daggermouth—a viral science-fiction romance that secured a seven-figure deal with Simon & Schuster—scored approximately 60% AI-written. The researchers used both AI detection scanning and linguistic analysis to identify rare, identical phrases appearing across multiple books that are more prevalent in AI-generated text than human writing, including duplicated sentences in Daggermouth that appear word-for-word in other self-published ebooks.
Daggermouth's potential AI involvement raises uncomfortable questions about whether AI fiction is improving in quality, having achieved mainstream success with viral TikTok promotion and best-seller status. Author H.M. Wolfe vehemently denies AI use, stating through her lawyer that she wrote the book herself and opposes generative AI in creative work. The study, which has not yet been peer-reviewed, highlights a growing challenge in the publishing industry as AI-generated ebooks flood Amazon's self-publishing platform.
- Creative industry faces an ethics and disclosure problem: The lack of AI content disclosure requirements creates unfair competition for human authors and raises reader trust concerns
Editorial Opinion
The rise of high-quality AI fiction represents a potential watershed moment for creative industries. If bestselling novels can be generated or substantially augmented with AI tools while evading detection and deceiving readers, the publishing industry urgently needs mandatory AI disclosure standards and improved detection methods. This story also reveals a deeper challenge: as AI writing quality converges with human writing, technical detection alone may become insufficient—the industry will need ethical frameworks and legal accountability to protect both author livelihoods and reader trust.
