Μake us preferred on Google Shortly after scoring a multimillion-dollar book deal for a debut author, the novelist’s agents pulled the plug. They could no longer support the project, they said, because they couldn’t verify that the book had been wholly written by their client.
“This raises so many questions about authorship and what AI means for this industry,” the agency, Europa Content, told the publishers in a July letter after allegations of AI use swirled around “Call Me, I’ll Hide the Body,” a crime novel by Jerry Falade. “But those are questions for another day.”
Yet with a new AI scandal engulfing publishing seemingly every month, it’s become more difficult to punt questions about its impact into some distant future. The spectacular implosions of big book deals over suspected AI use—and fears about who might be next—are forcing a reckoning over the nature of authorship, the relationship between writers and publishers and the industry’s long-term survival. But nobody can seem to agree who exactly is responsible for solving this problem, or even how much a problem it actually is.
Last year, 70 authors including Margaret Atwood and Jonathan Franzen signed an open letter to publishers voicing concerns about generative AI and asking them to make “a pledge that they will never release books that were created by machines.” Yet the Big Five—Simon & Schuster, Penguin Random House, HarperCollins, Hachette Book Group and Macmillan—have largely been reluctant to make sweeping pronouncements about a technology that is both potentially lucrative and increasingly ubiquitous. (Many of them are also involved in lawsuits against tech companies for training their models on copyrighted books.)
Instead, they find themselves running to put out fires when scandals over AI use become too reputationally damaging to ignore.
Falade, a Southern Methodist University graduate student, has denied the accusations about his novel, calling them racially motivated. “This has just been so terrible for me,” he said in an interview on Thursday. “I haven’t been able to sleep well, I haven’t been able to eat.”
In March, Hachette Book Group canceled the U.S. release of Mia Ballard’s book “Shy Girl,” a horror novel about a woman held captive by a man she met online and forced to live as his pet, amid allegations that the author had used AI in the writing process. Ballard, who originally self-published the novel before Hachette acquired the rights, has denied these claims, saying AI was used by an acquaintance who had edited the book. “Hachette remains committed to protecting original creative expression and storytelling,” the publisher said in a statement at the time.
Last month, the Atlantic wrote that “Daggermouth,” a self-published dystopian romance whose rights were snapped up by Simon & Schuster, and which has consistently topped bestseller lists, contained hallmarks of AI use. The piece cited a research paper from Stony Brook University professor Tuhin Chakrabarty that ran the book through the AI detection tool Pangram. The paper has not yet been peer-reviewed, and the author has denied the allegations, publicly saying she does not believe that generative AI belongs in the writing process.
“We do not believe conclusions about an author’s work should be drawn from AI-detection tools that have been shown to produce false positives,” said a spokesperson for Simon & Schuster. “H.M. Wolfe wrote ‘Daggermouth,’ and we stand behind her and her work.”
Pangram’s CEO, co-founder and self-proclaimed “slop janitor” Max Spero said the tool’s high rate of accuracy has been validated by multiple third-party studies.
“Why would Simon & Schuster drop it? Because apparently it already had a huge amount of success. So people were liking it,” said Chakrabarty. “There is this bigger angle about the commercial aspect of generative AI fiction, and then do the organizations really care about holding to that standard?”
Desperate to find hits in a time when most people are reading less and less, publishers often look to pick up self-published works that have already found an audience, a category where agents and editors say AI use is especially prevalent. But self-published authors are far from the only writers using these tools. In May, author Steven Rosenbaum, who is published by an imprint of BenBella Books, came under fire when the New York Times found he had included AI-hallucinated quotes in his book “The Future of Truth: How AI Reshapes Reality.” He was open about using LLMs as a resource in the book’s acknowledgments.
“Given the speed at which the world is moving and publishing is moving, it seems inevitable that more of this is going to happen,” said Paul Bogaards, a longtime book publicity and marketing executive, noting that publishers are asking “what safeguards do we need to have in place to protect ourselves from this happening again?”
An executive at one of the Big Five publishers said they made the decision not to add specific contractual language on AI, given the rapidly evolving landscape, and that norms differ across imprints and editors.
“AI is not a substitute for human creativity,” a Penguin Random House spokesperson said. “We see AI as a tool that can support certain publishing workflows, but not as a replacement for authors, illustrators, publishing professionals or the human intellect and judgment that shape our books.”
Andy Hunter, the founder and CEO of Bookshop.org, feels that debates over where the line is for individual authors obscure a much bigger issue: the deluge of slop books online.
“Probably 99% of these books are meant to trick customers into buying something that is either plagiarized or filled with misinformation or incredibly thin content that was generated by an LLM,” Hunter said.
Chakrabarty’s study also found that 20% of Amazon ebooks in their data set featured “substantial” AI assistance. The online retailer said it takes measures to remove content that violates its guidelines.
James Daunt, the CEO of Barnes & Noble, received criticism after saying on the “Today” show in May that if customers want to buy AI books, “then we will stock them.”
In an interview, he clarified his remarks. “What I prefaced it by saying is we don’t at the moment,” he said. “We go to great lengths not to sell them.” Barnes & Noble removes books from its online store if they are discovered to be AI-generated.
“The books that I want to publish are books that say something original and different from what’s out in the world already, and I don’t think AI is going to come up with those,” said Cindy Spiegel, co-CEO of independent publishing house Spiegel and Grau. But, “if something comes along that’s really phenomenal and feels really meaningful, and I get something out of it, then I think it’s worth having that as part of our repertoire” (although, she notes, these works should always contain a clear disclaimer).
Right now, public sentiment is still largely against using AI to generate text, although using it for research and outlining is becoming more acceptable. If it turns out that people are happy to read AI-generated or AI-assisted books, especially in genres like fantasy and sci-fi where readers churn through books at a rapid clip, publishers don’t want to cut off a possible revenue stream in a world where those are few and far between.
“The standards around it don’t exist right now,” said Amanda Annis, an agent at Trident Media Group. “There’s just so much gray area that hasn’t come out in the wash yet.”
In an attempt to combat the deluge of slop, the Authors Guild introduced a “human authored” certification that writers can add to their books after signing an attestation. Yet some feel that anything that relies on an honor system lacks the teeth necessary to combat such a far-reaching issue.
Publishing has always been an industry rooted in trust. Editors rely on authors to attest that their works are both accurate and original. The absence of institutional guardrails becomes a topic of conversation every time there is a plagiarism scandal or a memoir’s truthfulness comes under scrutiny, and many worry that AI will be yet another stress test that will further lay bare the industry’s inherent weaknesses.
Agents, who act as the intermediary between authors and publishers, have found themselves on the front lines. “I think that agents and editors are now being asked to essentially be police,” said Emma Dries, an agent with Triangle House Literary. She said her inbox has been inundated with more queries than ever, which she attributes to writers using LLMs to mass-spam agents, and has received submissions that seemed clearly AI-generated.
She recently turned down a nonfiction writer whose work she was excited about after they admitted to using AI frequently—not to write, they claimed, but to get inspiration. “I’m sure that person is going to find an agent and get published, and I don’t know if that’ll ever be a conversation,” she said, noting that the person already had a number of offers of representation.
“When I see something that’s honestly badly written, it’s such a breath of fresh air now,” said Angeline Rodriguez, an agent at WME, noting that some of the submissions she’s getting now “make ‘Fifty Shades of Grey’ look like Tolstoy.”
Some agents simply aren’t asking—and hoping it doesn’t come up. Others are taking a hard line, telling authors that they won’t represent anyone who uses AI at all, even for research. And others are using AI-detection software to vet manuscripts, although the accuracy of these tools remains a source of debate.
“There’s a lot of anxiety,” said Regina Brooks, CEO of Serendipity Literary Agency and president of the Association of American Literary Agents (AALA). “People are nervous that they’ve worked on a manuscript for a long time and then they may have pulled something from some research and they’re freaking out about whether this would be considered something that wasn’t copyrightable.”
The fact that AI-generated text can’t be copyrighted has made it one of the hard lines the industry has been able to enforce. Everything else, beyond copy-and-pasting sentences directly from an LLM into your manuscript, is a gray area.
“We’re all kind of trying to come up with guidance as we go, but it really does feel like the concerns are different this month than they were last month,” said Hannah Bowman, an agent at Liza Dawson Associates who chairs the AALA’s AI Special Committee.
Agents and editors know it’s only a matter of time before the next big AI scandal. “The only silver lining is, thank God, we’re still canceling these books,” as one AI-phobic agent put it. “I don’t know if that will continue to be the case.”
