Four years after the release of ChatGPT to the public, publishing is still coming to terms with the impact of AI. Today, the technology is mostly improving life in the back office, where publishers have built tools to automate everyday tasks, including developing metadata, producing royalty statements, proofreading, and providing customer service. In its 2025 Salary and Jobs Report, PW reported that 63% of industry professionals surveyed said their organizations were using AI. A forthcoming report on AI use by the Book Industry Study Group and BookNet Canada landed on the exact same figure.
Has AI use plateaued? Not exactly, but a more studious attitude has replaced the excitement of early adopters.
Keith Riegert, president of the Stable Book Group and CEO of Perfect Bound, has been a vocal advocate for the use of AI in publishing—despite calling himself a “hardcore techno-pessimist.” He tells PW, “I do this because you use AI or you die at this point.”
On the positive side, he describes how his team created an AI agent that pulls sales data from Amazon, e-book distributors, Faire, Hachette, and Shopify, and automatically generates monthly client statements. “We’ve gone from a hundred hours of work down to 15 minutes,” he explains. He’s also used AI to lay out a full book in InDesign, reformatting an older title into a new trim size, and says it did “a genuinely professional job.”
More broadly, his view has shifted from expecting AI to eventually run entire publishing companies autonomously to seeing it as infrastructure. “It’s really like the transition from typewriters to word processors, again, and this is the new Microsoft Word,” he notes. Like many in the industry, he maintains any use of AI must “start with a human and end with a human,” and he draws a clear line, avoiding AI for core creative work, such as selecting titles, designing covers, or creating illustrations—reserving it for developmental editing, formatting, and admin instead.
The Big Five have been relatively terse in their public comments about AI. In January 2025, Bertelsmann, the parent company of Penguin Random House, announced, for example, that it had deployed OpenAI’s ChatGPT Enterprise for all its employees and was using AI for marketing and to forecast demand. In March of this year, while discussing the appointment of Darryll Colthrust as Macmillan’s first chief technology officer, CEO Jon Yaged framed AI adoption as a strategic necessity, saying, “Technology is central to how we serve our authors, employees, and partners, and how we compete and grow in a rapidly evolving marketplace.”
What is known is that each of the Big Five is developing proprietary AI systems for internal use, as proven by the numerous AI-related job openings the companies have posted.
Growing Pains
Publishers have a strong incentive not to speak too openly about AI, as there is a risk of facing questions for which the industry does not yet have answers. Case in point: at PW’s U.S. Book Show in June, Authors Equity’s Madeline McIntosh joined Hachette Book Group CEO David Shelley and Sourcebooks publisher and CEO Dominique Raccah for an onstage interview with PW editorial director Jonathan Segura about AI. Shelley was in the hot seat and had to face questions about how the company failed to detect early on that Mia Ballard’s novel Shy Girl was written with the help of AI, with the company subsequently pulling it from shelves and canceling it. Asked how publishers can prevent this from happening again, Shelley was direct: “I honestly think anyone who thinks they know the answers here is lying or delusional, because we’re all trying to figure this out.”
All three said they are deploying AI in operations while drawing firm lines around creative work, and emphasized that they need to trust their authors not to use AI to write their books. Raccah warned the industry to be wary of starting an AI “witch hunt” and that too much reluctance would lead to lost opportunities for innovation. McIntosh said Authors Equity had yet to encounter any problems related to AI, but added, “I feel like every publisher is five minutes away from having an AI controversy.”
Then, as if on cue, came the controversies. First, in July, the Atlantic published a study that analyzed thousands of Kindle e-books using an AI-detection tool called Pangram. It singled out author H.M. Wolfe, accusing her of using AI to write parts of her novel Daggermouth, which was previously self-published and then bought and republished—as part of a seven-figure, two-book deal—by Simon & Schuster’s Scarlett Press imprint. S&S stood behind its author, and the book went on to become a bestseller.
In August, the novel Call Me, I’ll Hide the Body by Nigerian author Jerry Falade was pulled from submission after his agents, Marc Gerald at Europa Content and Sandy Hodgman at Hodgman Literary, became worried that Falade had written it with AI assistance. This happened after the book had already gone to auction, where it garnered numerous six- and seven-figure offers, including one that Falade accepted from Macmillan’s Minotaur imprint. For his part, Falade denies the accusation of having used AI, calling it racist, and has since secured David Vigliano as his new literary agent.
Smaller publishers, too, must cope with the challenge of identifying AI-generated text in books authors have submitted, and it isn’t always easy. At Microcosm, an independent publisher and distributor in Portland, Ore., co-owner and vice president Elly Blue says her team discovered an author’s undisclosed AI use almost by accident, when a freelance editor flagged what looked like plagiarism, only to find out it was AI-generated text. That discovery prompted Microcosm to run AI detection across its entire pipeline, which turned up several more cases, including two manuscripts fully generated from prompts, and one where an author had trained an LLM on his own past writing. “The book sounded like him, but it made no sense,” Blue recalls.
Confronting the authors had mixed results. “Some denied it outright and insisted, ‘That’s just what my voice sounds like,’ ” Blue says. “One, surprisingly, owned it and said he’d take the book elsewhere.”
These episodes led directly to Microcosm instituting a new contract clause that states that if an author’s undisclosed AI-generated text can’t be revised into something genuinely their own, Microcosm retains the rights and can publish the work itself, since the author didn’t actually write it.
Princeton University Press has also seen a rising number of allegations that its authors are using AI-generated text, according to CEO Christie Henry. “Trust with authors is foundational, so we rely on disclosures, contractual indemnities, and warranties, and we handle each allegation case by case.”
But Henry is less concerned about authors secretly using AI to write than about the time spent policing the situation. “It pulls us away from having focused time to think,” she says. PUP is currently vetting companies that promise to be able to certify that a given book was written by a human, including Created by Humans, Human Authored, and Verify My Writing.
Henry adds that the proliferation of plagiarized, copycat, and pirated books online, now supercharged by AI agents mass producing and reproducing content, is further draining resources, as it demands constant vigilance and takedown requests sent to Amazon, YouTube, and others. Riegert has a similar complaint.
Lawsuits vs. licensing
As publishers struggle with the limits of AI use internally, and the limits of AI detection in their submissions pipelines, they’re battling the tech giants over how their books are used to train AI models. The Association of American Publishers has been instrumental in supporting numerous lawsuits alleging copyright infringement by AI companies. Chief among these has been Bartz v. Anthropic, a case in which Anthropic agreed to pay $1.5 billion—not for copyright infringement but for using some 500,000 pirated books as source material.
Roughly 15 other author-driven lawsuits are underway against AI companies, some of which have been spearheaded by bold-faced names like Michael Chabon, George R.R. Martin, and Scott Turow. (Many are also parties to publisher suits.) In one instance, John Carreyrou and a group of six authors filed individual infringement claims against Anthropic, Google, Meta, OpenAI, Perplexity AI, and xAI seeking $150,000 per book, per company.
Meanwhile, publishers have filed roughly 10 coordinated suits of their own. The two most recent include a May suit against Meta from Cengage, Elsevier, Hachette Book Group, Macmillan, McGraw Hill, and Turow, and a July suit against Google from Cengage, Elsevier, Hachette, and Turow, both alleging widespread infringement in the training of the tech company’s AI models.
While the lawsuits play out, some publishers are building bridges. Google recently announced a partnership with Bloomsbury, De Gruyter Brill, Johns Hopkins University Press, Macmillan, O’Reilly Media, and Penguin Random House to enhance e-books from its Google Play store with AI, allowing readers to interact with the text of the books directly, ask questions and get answers, and develop tools such as quizzes and slideshows. The program, called Expert Intelligence, was praised by Mary Rasenberger, CEO of the Authors Guild, which has offered lengthy guidance for authors on how to handle AI provisions in contracts and has spearheaded several copyright lawsuits. She called the partnership “a positive step forward in the legitimized use of books in the AI ecosystem to the benefit of readers, publishers, and authors alike.”
Of even more significance is the growing number of licensing deals publishers are striking with AI companies. In 2024, HarperCollins became the first of the Big Five to publicly reveal that it had signed a deal to license its content for AI training. The three-year agreement paid $5,000 per participating title, split equally between HarperCollins and the authors. Wiley announced its first AI licensing deal in 2024 and has since made several more, including with Amazon, Anthropic, and Perplexity. For fiscal year 2026, it reported earning $49 million in revenue from the deals.
Cengage has meanwhile partnered with Amazon Web Services to build AI features inside Cengage-owned learning products, and Princeton University Press has announced a partnership with Cashmere.io, a Utah-based platform that licenses content for AI training.
“Licensing is a conversation that is happening,” Princeton’s Henry says. “It is ongoing.” She notes that much of the work of developing guidelines for equitable deals between publishing and Silicon Valley is being done by advocacy organizations like the AAP, the Authors Guild, and the International Publishers Association.
Whether AI companies will ever treat publishers as equal partners is another question altogether.
“Am I optimistic we’ll be invited to the table?” Henry asks. “No. There may be some irreconcilable differences we just have to face.”
She cites, among other things, a mismatch in the cultures of publishing and tech—the former having been around for several hundred years and the latter relatively new and so far exhibiting minimal regard for the written word, save for its utility as an AI training resource.
The fact is that publishing’s artistic, ethical, moral, and even commercial concerns may ultimately matter little to AI companies, many of which are individually valued at hundreds of billions of dollars. Money, after all, is power.
At least, Henry says, the direst early predictions about AI bringing a swift end to publishing haven’t come true. “Fundamentally, I don’t think AI can replace the publishing industry,” she notes. “People forget that we still have real agency in the decisions we make. We can still choose what to publish. We can choose what to buy. And, of most importance, we can still choose what to read.”
That’s all true. At least for now.
Read more from our AI in publishing feature.
Ingram’s AI Rollout Is a Success
An AI Company That Picks Up the Phone: PW Talks with Hannah Johnson
A version of this article appeared in the 09/14/2026 issue of Publishers Weekly under the headline: Publishing’s AI Reckoning