Most self-published authors can only dream of writing something as popular as Daggermouth. Set in a surveillance state dominated by masked elites, the novel traces the unlikely romance between the president’s son and the assassin hired to kill him. Since HM Wolfe uploaded Daggermouth on Amazon as a Kindle ebook, it has become very popular. “I’m done Daggermouthor did Daggermouth finish me?” TikTok book influencer poured out. In February, Simon & Schuster paid seven figures to acquire the publishing rights to the book and its sequel, which will be released next year.
Daggermouth It’s been a few months USA Todaybestseller list and ranked #1 in Amazon’s “science-fiction romance” category. It also tops an unofficial list, compiled by scholars who study AI writing. Among the thousands of e-books Amazon identified as containing “big AI text,” none stood out Daggermouth.
Until recently, AI fiction has been easy to dismiss as hype. On Amazon, a new book is about to be released three times from 2022 to 2025. (Perhaps not because humans suddenly committed themselves en masse to the written word.) clear copies of existing titles, which seem intent on tricking readers into buying them. Their sales figures are generally low; opinion, anger. A dubious number of them involve the main characters same names. Last year, readers of one fantasy novel i saw a line that seemed to quickly respond to asking the chatbot to write in the style of another writer. But if it is a favorite book Daggermouth were written by a chatbot, it would raise an awkward question: Can AI stories get … good?
In the new readingresearchers randomly selected more than 14,000 Kindle books and ran them through Pangram, an AI recognition tool. One in five returned with more than 25 percent of the written AI; Daggermouth it returned 60 percent, based on a primary data set the authors shared with me. Those numbers, they admit, only tell us so much. The researchers grouped together text that Pangram had flagged as “AI-generated” or “partially AI-assisted” (the latter of which could mean that the AI was used to edit or polish the human text, or vice versa). And while Pangram has a low rate of false positives, it is it is not perfect.
So the study’s authors, led by Tuhin Chakrabarty, a computer science professor at Stony Brook University, devised another check. They scanned the text of each book for “rare words”—strings of at least five words in a specific order—which also appear in other suspected AI books but almost never in those written by humans. Daggermouth they had many of them. For example, the following sentence appears near the end of a multi-page display of sex in Chapter 27: “They fell together, a mass of sweaty limbs and racing hearts.” The same sentence, word for word, also appears in another self-published e-book by a different author. Or consider this verse from Chapter 38 of Daggermouth: “His blow thundered in his ears, and silenced the murmuring of the multitude.” It’s also in an ebook by another author who has published over 400 different titles on Amazon over the past three years.
The study has not yet been peer-reviewed, and to be clear, there is no definitive proof that Wolfe used AI. “The comment that I used production AI to write Daggermouth absolutely not true,” Wolfe said in a statement sent by his attorney. “I wrote this book myself. I’ve been open about my opposition to production AI and what it’s doing to writers, artists, and the creative community. I don’t believe it’s in the writing process.” Referring to Pangram’s shortcomings, Wolfe added that “such accusations have real consequences for authors, and should not be made based on technology that is known to be unreliable.” Susannah Lawrence, spokeswoman for Daggermouththe 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 “rare words” which Daggermouth it is shared with other ebooks recently.
Chakrabarty, who researches AI writing, told me so Daggermouth he has “too many signs” of AI for him to buy that information. He cited Chapter 18, “Thank You,” as an example: His research team divided it into 12 pieces for Pangram to evaluate, and the tool classified all of the pieces as AI-generated. (In my spot check, the tool classified the first few paragraphs as human-written and the rest of the chapter as 100 percent AI.) I ran this with Mohit Iyyer, a professor of computer science at the University of Maryland who was not involved in the research. An AI language discovery expert, Iyyer told me that even counting Pangram errors, “it’s almost statistically impossible” for a human-written book to register as 60 percent AI. He added that the broader research approach, which links AI writing to ebook sales, is good.
Whether or not you believe Wolfe deserves the benefit of the doubt, the same questions about how Daggermouth was written to be used for a growing number of other hits on Amazon. Chakrabarty’s research found that books written with “massive” AI support make up 20 percent of Amazon’s e-book inventory, 12 percent of sales, and 10 percent of best sellers in their genre.
Not yet Daggermouth it stands for more than just its popularity. The second best-selling e-book in the data set is Ink Athenaand Nicole Fox, an author who has 418 titles listed on Amazon. Daggermouthby contrast, it appears to be one of only two novels to Wolfe’s name in 2025, and its sequel won’t come out until 2027 — an indication that Wolfe’s process may involve a significant amount of genuine human craftsmanship. Iyyer said that using AI in a way that results in a novel that is fun to read takes “a lot of effort on the part of a human writer.”
Another difference: Jennifer Prokop, who reviews books for Kirkus and co-hosts Destined Spousespopular podcast about romance novels, told me that people in his world keep talking about it Daggermouth. “Nicole Fox, I’ve never heard a single soul talk about her,” he said. (Fox did not respond to a request for comment.) Prokop sees mechanical prose increasingly in self-published e-books, where its main platform is Amazon’s Kindle Unlimited subscription service. Just because an ebook is charting on Kindle doesn’t mean people are buying it. Grifters can now use AI to hand out dozens of titles and then use all sorts of tricks to attract readers, like buying hundreds of fake five-star reviews and paying click farms to milk their metrics. “If you can get 10,000 people to read 10 pages of your book, and you have 300 books to give away,” he said, you can make real money. “But you don’t really write romance that people want to read.” (Amazon did not respond to a request for comment.)
Like Daggermouth is AI, then maybe it’s an outlier—a book that captured the full wave of social media buzz. (A series of controversies surrounding the book, including accusations of the author’s use of AI and the retreat of the cover art for a special hardcover edition, seemed to embolden his fans and bring more attention.) The notion that AI writing requires a perfect combination of events to top the charts is a comforting thought. But it may not be true. “Everyone thinks they can tell AI text,” Christine Larson, a journalism professor at the University of Colorado at Boulder who studies technology’s impact on creative work, told me. “But I don’t think we can say much.” Privately, several writers and editors have told him they use technology to speed up or improve their results. “I totally agree with the idea that a human working with AI can write something that’s compelling,” he said. “I have no doubt that we will see more in the future.”
If he’s right, then the question isn’t whether we’ll see AI marketers. It’s about whether anyone involved with them will agree. Coral Hart, a South African author who uses AI to help her publish hundreds of books a year under different pen names, told me she works hard to hide her pseudonyms after receiving death threats and thousands of one-star reviews following New York Times a story about him. Hart claimed that he has had successful writers join the classes he teaches on AI writing, though he declined to name them. “Until they stop the harassment, the witch hunts, and get rid of them, I don’t think the writers will be open, and I don’t blame them,” Hart said. “Of course I’m not going.”
The taboo against the use of AI in literature is understandable: Among other things, it’s wrong to take credit for a book you didn’t write yourself. Authors suspected of employing AI might work if there was any way to tell for sure—which there isn’t. The shaming ritual places the moral burden on individual authors rather than the large corporations responsible for the system they are trying to make a living from. It also tends to blur any distinction between pure AI sleight of hand and human writing that is aided by AI. The result is a state of affairs that feels unreliable: AI writing can be very profitable, but only if it is kept secret. Finally, something has to give.
