Shy Girl, a horror novel by Mia Ballard, was one of those buzzy books that leapt from self-published prominence into full-on trade publication. Until yesterday, that is, when publisher Hachette pulled the book from the UK market and canceled plans to bring it to the US.
The move came after a New York Times investigation suggested that AI had been used in significant parts of the work.
“If it isn’t AI, she’s a terrible writer”
Shy Girl was self-published in 2025 and quickly found an audience on social media. The novel follows a depressed, OCD woman named Gia who, down on her luck, encounters a “sugar daddy” who pays off her debts. All she has to do? Live as his literal pet. Eventually, of course, living like an animal makes her into an animal, and things apparently get nasty.
Creepy. And the prose? “I’m obsessed with the way Mia Ballard writes,” said one reviewer on Goodreads.
Not everyone thought the book was good, though, or even well-written. Another reviewer on the site called the book “absolute f—ing garbage. overwritten, repetitive, poorly executed, atrocious formatting. nothing to do with actual feminine rage and revenge.”
Soon, the questions moved beyond the literary. Had the book really been “written” at all? Complaints started to surface that the prose sounded, at least in places, like chatbot writing.
In January 2026, someone claiming to be a long-time book editor posted a long Reddit thread claiming that the novel had all the hallmarks of AI lit. “If so, I find it repulsive that it has been picked up and published by the second largest publishing company, at least in the UK,” said the Reddit post. “If it isn’t AI, she’s a terrible writer. Her writing is truly indistinguishable from an LLM.”
Then a two-and-a-half hour (!) YouTube video dropped, making the same claims. It garnered 1.2 million views.
Even AI detection companies like Pangram got in on the action, claiming that the book had the hallmarks of being largely AI-generated.
Still, Hachette appeared to be moving forward with plans for a US release later this year.
Yesterday, The New York Times published its own investigation, in which it “analyzed passages from the novel using several AI detection tools and found recurring patterns characteristic of AI generated text, like gaps in logic, excessive use of melodramatic adjectives and an overreliance on the rule of three.”
That did it. Hachette pulled the book in the UK and canceled its upcoming US debut. Late last night, the Times received a comment from Ballard, the author, denying that she had used AI to write the novel. And yet, Ballard added, it was possible that a friend who helped edit the book did use AI.
“This controversy has changed my life in many ways and my mental health is at an all-time low and my name is ruined for something I didn’t even personally do,” Ballard added and claimed that she was pursuing legal action.
But what if it’s “good enough”?
This is one of the first major AI controversies to hit the world of traditional trade publishers, where the old-school gatekeepers still largely ban AI—at least for drafting. (Outlining, edit feedback, plot suggestions… all of these are far murkier.)
Whatever actually happened in this situation, publishing is likely to see similar disruptive patterns to those roiling the music industry right now, where tools like Suno are increasingly used to crank out songwriting demos and even (at least on places like Spotify) fully AI-produced music. Many artists and even distributors such as LANDR resist such AI use, but plenty of ordinary people don’t care. Their view seems to be: If the music sounds good—or good enough—what difference does it make where it came from? And, frankly, how different is super-glossy pop from the kind of thing Suno turns out?
In the case of Shy Girl, despite numerous claims that AI writing sucks and that it can be easily identified, plenty of readers enjoyed the book and even promoted it online. That may both terrify and horrify actual writers, but it remains a reality they’ll need to face.
