Most self-published authors can only dream of writing something as popular as Daggermouth. Set in a surveillance state ruled by a masked elite, the novel traces the unlikely romance between the president’s son and an assassin hired to kill him. Since H. M. Wolfe uploaded Daggermouth to Amazon as a Kindle ebook, it has been a viral hit. “Did I finish Daggermouth, or did Daggermouth finish me?” a TikTok book influencer gushed. In February, Simon & Schuster paid seven figures to secure the publishing rights to the book and its sequel, which is due out next year.Daggermouth has spent months on USA Today’s best-seller list and is ranked No. 1 in Amazon’s “science-fiction romance” genre. It also tops a less prestigious list, one compiled by academics studying AI writing. Among thousands of self-published Amazon ebooks that they identified as containing “substantial AI text,” none was more popular than Daggermouth.Until recently, AI fiction has been easy to dismiss as slop. On Amazon, new-book releases nearly tripled from 2022 to 2025. (That’s probably not because humans suddenly dedicated themselves en masse to the written word.) Many are blatant copycats of existing titles, seemingly intended to trick readers into buying them. Their sales figures are generally meager; the reviews, indignant. A suspicious number of them feature protagonists with the same names. Last year, readers of one fantasy novel noticed a line that seemed to be responding to a prompt asking a chatbot to write in the style of another author. But if a book as beloved as Daggermouth were written with a chatbot, it would raise an awkward question: Could AI fiction be getting … good?In a new study, researchers randomly selected more than 14,000 Kindle ebooks and ran them through Pangram, an AI-detection tool. One in five came back as more than 25 percent AI-written; Daggermouth returned 60 percent, according to an underlying data set that the authors shared with me. Those numbers, they acknowledge, tell us only so much. The researchers grouped together text that Pangram had flagged as either “AI-generated” or “moderately AI-assisted” (the latter of which can mean that AI was used to edit or polish human writing, or vice versa). And although Pangram has a low rate of false positives, it is not perfect.[Read: America has a Pangram problem]So the study’s authors, led by Tuhin Chakrabarty, a professor of computer science at Stony Brook University, devised another check. They scanned each book’s text for “rare expressions”—strings of at least five words in a specific order—that also appear in other suspected AI books but almost never in human-written ones. Daggermouth had many of them. For instance, the following sentence appears near the end of a multipage sex scene in Chapter 27: “They collapsed together, a tangle of sweat-slicked limbs and racing hearts.” That same sentence, word for word, also shows up in another self-published ebook by a different author. Or consider this line from Chapter 38 of Daggermouth: “Her pulse thundered in her ears, drowning out the murmurs of the crowd.” It’s also in an ebook by another author who has published more than 400 different titles on Amazon in the past three years.The study has yet to be peer-reviewed, and to be clear, there is no definitive proof that Wolfe used AI. “The suggestion that I used generative AI to write Daggermouth is wholly untrue,” Wolfe said in a statement sent by her lawyer. “I wrote this book myself. I have been outspoken about my opposition to generative AI and what it’s doing to writers, artists, and the creative community. I don’t believe it belongs in the writing process.” Referencing Pangram’s limitations, Wolfe added that “accusations like these have real consequences for authors, and they should not be made on the basis of technology that is known to be unreliable.” Susannah Lawrence, a spokesperson for Daggermouth’s publisher, Simon & Schuster, told me in an email that the company stands behind the book, noting that it went through “the same editorial and production process as our other published titles.” Neither Wolfe nor Lawrence addressed my questions about the “rare expressions” that Daggermouth shares with other recent ebooks.Chakrabarty, who researches AI writing, told me that Daggermouth has “too many telltale signs” of AI for him to buy that explanation. He cited Chapter 18, “Thank You,” as an example: His research team broke it into 12 chunks for Pangram to evaluate, and the tool classified all parts as AI-generated. (In my own spot checks, the tool classified the first few paragraphs as human-written and the rest of the chapter as 100 percent AI.) I ran this by Mohit Iyyer, a professor of computer science at the University of Maryland who wasn’t involved with the study. An expert on AI-language detection, Iyyer told me that even accounting for Pangram’s fallibility, it’s “almost statistically impossible” for an entirely human-written book to register as 60 percent AI. He added that the study’s broader methodology, which connects AI writing to ebook sales, is sound.Whether or not you believe that Wolfe deserves the benefit of the doubt, the same questions about how Daggermouth was written apply to a growing number of other hits on Amazon. Chakrabarty’s study found that books written with “substantial” AI assistance made up 20 percent of Amazon’s ebook catalog, 12 percent of sales, and 10 percent of best sellers in their genre.Yet Daggermouth stands out for more than just its popularity. The second-best-selling ebook in the data set is Inked Athena, by Nicole Fox, an author who has 418 titles listed on Amazon. Daggermouth, in contrast, appears to be one of just two novels to Wolfe’s name in 2025, and its sequel is not due out until 2027—an indication that Wolfe’s process likely involves significant amounts of genuine human craft. Iyyer said that using AI in a way that results in a novel that’s actually enjoyable to read takes “a huge amount of effort by a human author.”[Read: The future of writing is a lot like hip-hop]Another difference: Jennifer Prokop, who reviews books for Kirkus and co-hosts Fated Mates, a popular podcast about romance novels, told me that people in her world keep talking about Daggermouth. “Nicole Fox, I’ve never heard a single soul talk about her,” she said. (Fox did not respond to a request for comment.) Prokop sees mechanical prose cropping up mostly in self-published ebooks, for which the dominant platform is Amazon’s Kindle Unlimited subscription service. Just because an ebook is charting on Kindle doesn’t mean people are devouring it. Grifters can now use AI to crank out titles by the dozen and then use all kinds of tricks to entice readers, such as buying hundreds of bogus five-star reviews and paying click farms to juice their metrics. “If you can get 10,000 people to read 10 pages of your book, and you have 300 books you’re putting out,” she said, you can make real money. “But you’re not actually writing romances people want to read.” (Amazon did not respond to a request for comment.)If Daggermouth is AI, then maybe it’s just an outlier—a book that caught the perfect wave of social-media hype. (A series of controversies surrounding the book, including accusations of the author’s AI use and a backlash to the cover art for a special hardcover version, seemed to galvanize its fans and bring it even more attention.) The notion that AI writing requires a perfect confluence of events to top the charts is a comforting thought. But it might not be true. “Everyone thinks they can tell AI writing,” Christine Larson, a journalism professor at the University of Colorado at Boulder who studies the effect of technology on creative work, told me. “But I don’t think we can tell so much.” Privately, several authors and editors have told her that they’re using the technology to accelerate or improve their output. “I totally buy into the idea that a human working with AI can write something that is compelling,” she said. “I have absolutely no doubt that we will see more of this in the future.”If she’s right, then the question is not about whether we’ll see AI best sellers. It’s about whether anyone involved with them will admit it. Coral Hart, a South African author who uses AI to help her publish hundreds of books a year under different pen names, told me that she works hard to keep her pseudonyms secret after she received death threats and thousands of one-star reviews following a New York Times story about her. Hart claimed that she has had successful authors join the classes that she teaches on AI writing, though she declined to name them. “Until they stop the bullying, the witch hunts, and the ousting, I don’t think authors are going to be transparent, and I don’t blame them,” Hart said. “I’m certainly not going to.”The taboo against the use of AI in literature is understandable: Among other things, it’s wrong to take credit for a book that you didn’t write yourself. Shunning authors suspected of employing AI might work if there were any way to tell for sure—which there isn’t. That shaming ritual puts the moral burden on individual authors instead of on the mammoth corporations responsible for the system in which they’re trying to eke out a livelihood. It also tends to collapse any distinction between pure AI slop and human writing that is assisted by AI. The upshot is a state of affairs that feels untenable: AI writing can be lucrative, but only if it’s kept secret. Eventually, something’s got to give.