Should AI be allowed in books?
Mia Ballard is a writer in California. Her official bio says that she writes horror stories focused on feminine rage. Horror stories give me nightmares, and my life is mercifully free of feminine rage. So I probably wouldn’t know anything about Mia Ballard, except that this year she’s the center of the biggest story in the publishing industry.
Her book, “Shy Girl,” is about a woman held captive and forced to live as a man’s “pet.” Ballard self-published the book in February 2025 on Amazon. It was available as an e-book and in print-on-demand. (That means, Ballard herself didn’t print any copies up front. Instead, Amazon printed a single copy whenever a customer ordered a book.)
The book became a success by self-publishing standards. It got about 5,000 ratings on Goodreads and earned some buzz on TikTok—so much so that it caught the eye of an editor at Hachette, one of the “big five” English-language publishers. Hachette acquired the rights to “Shy Girl” and released it via traditional channels in the United Kingdom in late 2025. It was scheduled to come out in print in the United States in the spring of 2026.
But not all of the online buzz was positive. One YouTuber published an almost-three-hour video takedown titled, “i’m pretty sure this book is ai slop.” Reddit threads analyzed the syntax and concluded it had all the hallmarks of ChatGPT writing. One character’s voice was, “a blade cutting through the sweetness”; another’s was “a blade slicing through the air”; a third’s was “a blade wrapped in silk.” Saying one person’s voice is a blade is a stretch: do it three times, and that’s just bad writing.
Enter Panagram, an AI detection site. Panagram claims that it can identify whether a piece of writing is created by humans or artificial intelligence. Early in 2026, after the book’s publication in the U.K., the president of Panagram said that “Shy Girl” was 78.3 percent AI-generated. Still, these rumors and discussions were relatively low profile, confined to book- and publishing-related forums.
But then in mid-March 2026, the New York Times was working on a story about the online buzz about AI and “Shy Girl.” They reached out to Hachette, the publisher, for comment. And that same day, Hachette decided to pull the book from publication. That means they will never release the book in the United States. They removed the book from store shelves in the U.K., and the ebook is no longer available for sale.
In a statement, Hachette says that it is “committed to protecting original creative expression and storytelling” and that all writers are required to attest that their work is original to them. They are supposed to disclose whether and how they used AI in the writing process.
Ballard, the author, denies using AI during the writing process. But she did acknowledge that her freelance editor used AI, and so some AI-generated phrases may have made their way into her final draft. She thinks the book’s cancellation, and the subsequent hit to her reputation, are unfair.
Nobody comes out of this story looking good. Start with the author. First, her book is full of the tells of AI-generated text. She says a freelance editor rewrote large parts of the text. Very well, but what does that say about her as an author—that she would allow a friend to rewrite her own book with ChatGPT?
Hachette, the publisher, seems like the victim, the wounded party here. But that doesn’t stand up to scrutiny. Hachette surely knew about the online buzz earlier. The company says it pulled the book after a “lengthy review”—but that coincidentally happened the day a big newspaper asked them for comment. The time for a lengthy review of a book’s text would be before printing, not after.
Panagram, the AI detection service, might seem innocent. They’re safeguarding the public from inadvertent exposure to AI slop. But keep in mind, Panagram has a service to sell. It benefits them to make accusations because it keeps them in the news. Their claim that the book was 78.4 percent AI generated is oddly precise. And it’s worth mentioning this irony. Panagram, which protects the public from harmful artificial intelligence, is itself an AI product.
Jeff’s take
I signed up for a free account at Panagram and decided to test it out. I know that Plain English episodes are human generated. But I still do use AI the way writers are supposed to: for fact checking, grammar checks, and advice on organization, flow, usage, things like that. I never copy text from Claude into an episode draft. Still, if Panagram said my writing was AI-generated, then that would be a comment on the quality of my work more than anything else.
You can imagine my relief when Panagram said with high to medium confidence that Plain English is 100 percent human generated.
Anyway, here’s my view. It shouldn’t matter if a word, a sentence, a paragraph, or 78.4 percent of something is AI generated. It should matter if it’s good. “Shy Girl” wasn’t to my taste. But if it makes people happy, then why shouldn’t they be able to buy it? Cancelling this book is like a blade sheathed in a leather-bound first edition.
Great stories make learning English fun
