For months, speculation has been building online that a buzzy horror novel, “Shy Girl,” was written with the help of A.I.
The novel, about a desperate young woman who is held hostage by a man she met online and forced to live as his pet, was self-published in February 2025. The book quickly found an audience among horror fans, and Hachette published it in the United Kingdom last fall and planned to release it in the United States this spring, billing it as “an unapologetic, visceral revenge horror novel.”
Earlier this year, Max Spero, the founder and chief executive of Pangram, an A.I. detection program, heard of the claims about “Shy Girl” and decided to run a test of the full text. Its results indicated that the book was 78 percent A.I. generated.
“I’m very confident that this is largely A.I. generated, or very heavily A.I. assisted,” said Spero, who posted his research on X in January.
In the months since “Shy Girl” was released in Britain, more readers voiced their suspicions online that the writer relied on A.I., citing nonsensical metaphors and odd, repetitive phrasing. As a chorus of allegations built online in late January that the novel was A.I. generated, Hachette stayed silent.
In response to questions from The New York Times about the A.I. allegations against “Shy Girl,” Hachette told The Times that its imprint Orbit has canceled plans to release the novel in the United States and that Hachette will discontinue its U.K. edition.
The author of “Shy Girl,” Mia Ballard, who according to her author bio writes poetry and lives in Northern California, has very little social media presence, and doesn’t appear to have addressed the allegations of A.I. use on her feeds. Reached by email on Thursday, Ballard said she was not prepared to comment.
The decision to cancel the publication came after a lengthy and thorough analysis, Hachette’s spokeswoman said, noting that the company values human creativity and requires authors to attest that their work is original. Hachette also asks its authors to disclose whether they are using A.I. to the company.
“Shy Girl” appears to be the first commercial novel from a major publishing house to be pulled over evidence of A.I. use. Its cancellation is a sign that A.I. writing is not only appearing in cheap self-published e-books that are flooding Amazon but is seeping into even traditionally published fiction.
The stunning fact that “Shy Girl” got so far into the editorial process, and was even released in the U.K. before publishers thoroughly investigated the claims of A.I. use, is a sign of how unprepared many in the book world are to deal with the rise of A.I. It also signals the dawn of an uncertain new era for the book world, as editors and readers alike are increasingly left wondering whether the prose they are reading was written by a human or a machine.
Few publishers or editors would talk on the record about how they are handling A.I. because its writing uses are so divisive and ethically murky. But some publishing executives worry that there is little that can be done to stop the A.I. incursion, especially as the technology rapidly becomes more sophisticated.
“It’s like with plagiarism — you’re at the mercy of the author,” said Morgan Entrekin, publisher of Grove Atlantic. “We have to have confidence in our partners.”
For now, the most obvious disruptions from A.I. are hitting the self-publishing sphere, where authors say the ecosystem has been flooded with A.I. slop. But some in the industry believe that it’s only a matter of time before more books written with A.I. slip past editors at major houses. The technology has become increasingly widespread — as has the practice of picking up self-published books and rereleasing them through traditional imprints.
“It’s not merely inevitable,” said Thad McIlroy, a publishing industry consultant who has urged publishers to clarify their policies around the technology. “We’re in the midst of it.”
After McIlroy learned of the allegations about “Shy Girl” from an employee at Pangram, he got a copy of the book and requested reports from Pangram and two other A.I. detection programs, GPT Zero and Originality.ai. All three found the text likely to be largely or partly generated by A.I., with Pangram’s report flagging certain phrases — including “the pause feels like a knife in my chest, sharp and unyielding,” and later, “I press the phone to my lips, the screen cool and unyielding” — as bearing the hallmarks of chatbot writing.
It’s nearly impossible to gauge how much A.I. writing is getting published, but there’s evidence that the technology has led to a surge of books. Last year, more than 3.5 million books were self-published, up from 2.5 million in 2024, according to Bowker, which collects book industry data. Traditional publishers released more than 642,000 books last year.
Tuhin Chakrabarty, a professor of computer science at Stony Brook University, used Pangram to check more than 14,000 self-published novels on Amazon for A.I. writing. The program found that nearly 20 percent of the novels had been substantially written by A.I. Looking mostly at novels released between 2024 and 2025, Chakrabarty saw a 41 percent jump year-over-year in how many novels in his random sample contained a large amount of A.I. generated text, he said.
A.I. detectors sometimes mistakenly flag human writing as computer generated. Still, Chakrabarty said he was confident that Pangram was picking up chatbot language. The program was built to detect linguistic patterns that are frequently used by large language models like ChatGPT and Gemini, and has a false positive rate of around one in 10,000, Spero said. It’s also designed to catch human efforts to cover up A.I. use through editing.
After compiling a list of self-published novels that Pangram showed to be heavily A.I. generated, Chakrabarty zeroed in on books that were resonating with readers, based on their number of Goodreads ratings and the average number of stars.
When he ranked the books according to the highest number of reviews, “Shy Girl” was among the most highly and widely rated, with more than 4,840 ratings and an average of 3.5 stars.
The first review on the page, however, awards it one star, from a reader who wrote, “I am quite certain that this was written by ChatGPT.”
Many publishers don’t explicitly prohibit authors from using A.I. in their book contracts. Instead, they rely on longstanding contractual clauses that require writers to affirm that their work is “original,” which many people in the book business now interpret as effectively banning the use of A.I. for text or image creation.
Publishers are also wary of A.I. content because currently, A.I.-generated text and art can’t be protected by copyright. Still, given the widespread uses for A.I. during research, outlining and other parts of the writing process, there’s little clarity on what constitutes its appropriate use. Many in the industry worry that publishers are leaving themselves vulnerable to scammers — or even writers who believe their A.I. use doesn’t cross any lines.
One problem in regulating authors’ A.I. use is that most corporate publishing houses don’t want to ban it outright. Editors recognize that authors use A.I. in a range of ways short of writing with it. And publishing executives want to ensure that their employees can use the technology for tasks like creating marketing copy, audio narration and translation.
The fact that publishing companies generally haven’t drawn a hard line around A.I. use is sowing confusion about what is permissible. Could a novelist ask A.I. to suggest plot twists, propose an alternate ending or polish a draft and still claim it as original work? At what point does the work stop being human?
Widespread suspicions around authors’ A.I. use has put publishers in a precarious position. Many still see traditional publishing as the only remaining fortress for original fiction that is handpicked and polished by discerning editors. If A.I. is capable of producing gripping fiction, and readers and editors are unable to discern its origins, it could erode publishers’ status as literary tastemakers.
“It’s a real problem, and we do have to find some guardrails,” said Mary Rasenberger, the chief executive of the Authors Guild, which is leading a class-action copyright lawsuit on behalf of authors against OpenAI and Microsoft, charging that ChatGPT was illegally trained on writers’ copyrighted works. (The New York Times has sued OpenAI, ChatGPT’s maker, and its partner Microsoft for copyright infringement of news content related to A.I. systems. Both companies have denied those claims.)
“There are publishers and authors who think the quality of A.I. is at a level that it’s not going to compete with them, and I don’t think that’s the case with the new large language models,” Rasenberger added.
Writing with A.I. remains wildly divisive among authors and readers. Some see it as a form of cheating, particularly if readers don’t realize that the book they’re reading includes passages generated by a chatbot. To others, it feels like theft: Many generative A.I. programs were trained on unlicensed copies of authors’ copyrighted works.
Seth Fishman, a literary agent, said the authors he represents are largely opposed to using A.I. for writing.
“For authors, this is not just a technology, it’s a moral issue,” he said. “Authors feel their work has been stolen.”
Even before “Shy Girl,” rumors circulated in the book world about near misses as publishers caught A.I. use before a book was released.
In one case, an editor at a major publishing house, who spoke on condition of anonymity because the editorial process is private, questioned a writer about why some passages in his latest book were so flat and bland. The author acknowledged that he had used A.I. for revisions.
Another publishing house found that a book it had acquired featured A.I. assisted prose and pulled it from publication, according to an employee who declined to elaborate or identify the author or the imprint, citing the confidential nature of publishing contacts.
There’s little consensus on what should be done to screen books for undisclosed A.I. text. Some agents and editors argue that publishers need to explicitly state their expectations around A.I., to prevent confusion or even fraud by authors who are not transparent about their reliance on the technology.
Penguin Random House, the largest publishing company in the United States, has created guidelines to set parameters around A.I. use for authors and illustrators that echo its contractual clauses stipulating originality. Representatives of other major publishing companies, including HarperCollins, Simon & Schuster and Macmillan, declined to elaborate on their A.I. policies apart from noting those originality clauses in their publishing agreements — although a spokeswoman added that Macmillan evaluates questions around A.I. use on a case-by-case basis.
Still, some worry that the ambiguity surrounding A.I., and the stigma the technology carries in the literary world, makes it more likely that writers might not be transparent about it.
“If it’s hush-hush, if you don’t want to talk about it, people are going to misuse it,” Chakrabarty said. “The shame around A.I. is causing more harm than help.”
Susan C. Beachy and Kitty Bennett contributed research.
Alexandra Alter writes about books, publishing and the literary world for The Times.
The post A.I. Is Writing Fiction. Publishers Are Unprepared. appeared first on New York Times.
