A self-published romance novel that rocketed to viral fame and landed a seven-figure deal with Simon & Schuster contains roughly 60 percent AI-generated text, according to a new academic study. The finding blurs the line between human and machine authorship as AI-generated books flood online marketplaces - and some are starting to sell.
The book is Daggermouth, written by H. M. Wolfe. Set in a surveillance state ruled by a masked elite, it follows the romance between a president's son and the assassin hired to kill him. After Wolfe uploaded it to Amazon as a Kindle ebook, it caught fire on TikTok. "Did I finish Daggermouth, or did Daggermouth finish me?" one book influencer said. Simon & Schuster acquired the rights in February for seven figures, and the novel has spent months on USA Today's best-seller list. It ranks No. 1 in Amazon's "science-fiction romance" category.
The study that identified it
Researchers led by Tuhin Chakrabarty, a computer science professor at Stony Brook University, randomly selected more than 14,000 Kindle ebooks and ran them through Pangram, an AI-detection tool. One in five books returned a result of more than 25 percent AI-written. Daggermouth came back at 60 percent, according to the underlying data set the authors shared. The researchers grouped together text flagged as either "AI-generated" or "moderately AI-assisted" - the latter meaning AI was used to edit or polish human writing, or the reverse. Pangram has a low false-positive rate, they said, but it is not perfect.
So the team added a second check. They scanned each book for rare expressions - strings of at least five words in a specific order - that appear in other suspected AI books but almost never in human-written ones. Daggermouth had many of them. One sentence, appearing near the end of a multipage sex scene in Chapter 27, reads: "They collapsed together, a tangle of sweat-slicked limbs and..."
AI fiction is no longer easy to dismiss
Until recently, AI-generated fiction was widely written off as slop. New-book releases on Amazon nearly tripled from 2022 to 2025, a surge unlikely to reflect a sudden mass dedication to writing. Many of these books are blatant copycats designed to trick readers. Their sales figures are generally meager, and reviews tend toward indignation. Some contain telltale artifacts - last year, readers of one fantasy novel found a line that seemed to respond directly to a chatbot prompt asking it to write in another author's style.
But Daggermouth changes the calculus. A book that readers genuinely love, one that commands a seven-figure advance and tops best-seller lists, turns out to have been produced with heavy AI involvement. It raises a question the publishing industry cannot ignore: Is AI fiction getting good enough to compete with human authors?
Why this matters for writers
The success of Daggermouth does not mean AI is replacing novelists. It does show that AI-assisted fiction can clear thresholds once reserved for human-only work: reader enthusiasm, critical attention, and major publishing money. Writers who understand how these tools work - and how readers and publishers respond to them - will be better positioned to navigate a market where the line between human and machine authorship keeps shifting. For those building skills in AI for Writers, the study offers a concrete benchmark: AI-generated text now appears in roughly one-fifth of new Kindle ebooks, and in at least one case, it has helped produce a genuine hit. Learning to use these tools deliberately through Prompt Engineering is becoming a practical skill rather than a theoretical exercise.
Your membership also unlocks:
