Publishers Cancel Books Over AI While Using It in Secret

A publisher can cancel your book for using AI, then write the press release about it with AI.

That’s not a hypothetical anymore. All year, the big American publishers have been playing defender of human writing. They’ve sued over copyright. They’ve pulled novels from authors suspected of letting AI help. Hachette canceled the American release of a horror novel. A literary agency withdrew its support from a crime novel it had already sold.

Meanwhile, behind the office doors, the same companies have been buying AI licenses by the box.

That’s the finding of a WIRED investigation published this month, built on interviews with more than two dozen publishing workers who spoke anonymously to protect their jobs. According to those workers, at least three of the Big Five, HarperCollins, Simon & Schuster and Hachette, are using AI tools across their operations without public disclosure or author consent.

I use AI every day as a digital assistant that helps me write, and I still do the writing. I don’t care what tool a publicist uses to draft a pitch. I care a lot about a company that punishes authors for something it does itself, and hides it.

Are book publishers using AI?

According to WIRED’s sources, yes, and in more places than most authors would guess.

Staff described AI being used to write publicity copy and back-cover descriptions, draft emails to literary agents, generate marketing materials and contribute to cover art. One HarperCollins division was described as fully committed, down to AI-generated marketing videos. Publicity copy came up again and again as the most common use. That’s the text readers see on Amazon, on Goodreads and on the back of the book itself.

At HarperCollins, senior leaders reportedly bought licenses for Anthropic’s Claude, then for ChatGPT and Jasper, a platform sold on running marketing workflows from start to finish. Staff say leadership didn’t know how to put the tools to work, so employees were drafted as “AI Champions” and sent into monthly brainstorming sessions to come up with uses. When some raised legal, ethical and environmental concerns, they say the person running the session brushed them off as part of doing business.

Simon & Schuster held a contest in 2024 for the best AI use cases, with a $10,000 prize, according to staff. Employees there also attended a workshop with OpenAI, where one suggested use was writing emails to literary agents. Staff say the company has since pulled back on some of the push after employees kept complaining at town halls.

Literary agents noticed before anyone told them. Several began to suspect that some editors’ rejection letters were coming out of an AI tool, and when they compared notes, the letters shared the same habits. Staff at all three houses told WIRED they believed the agents were right.

What do the publishers say about using AI?

Two of them answered, and their answers are carefully drawn.

Hachette told WIRED it supports AI for operational purposes that help authors’ books reach more readers. It said it doesn’t support creative uses, including communicating with authors and partners. Simon & Schuster’s global chief marketing and communications officer, Wibke Grutjen, said employees have access to a limited set of approved enterprise AI tools. She said using them isn’t required and unvetted tools are prohibited. HarperCollins didn’t respond.

Notice the gap. Hachette draws a line at communicating with authors and partners. Its own staff describe emails to agents as one of the growing uses across the industry. Either somebody inside isn’t following the policy, or the policy was written for the press and not for the people doing the work.

Why is it hypocritical for publishers to cancel books over AI?

Because the standard only runs one direction.

An author accused of using AI gets a detection score attached to her name and a canceled contract. Editors told WIRED that the canceled books that make the news are only a few of them. Nearly every imprint, they said, has a quiet story of a respected author whose book was dropped. There’s no hearing, no published rule and no way to prove a negative. I wrote about how broken that process is in Publishing Is Cancelling Books Over AI, and I called it cowardice then.

Now look at the other side of the desk. The same companies let AI write the copy that sells those books, the emails that reject them and some of the art on their covers, and they don’t tell anyone. No author got a disclosure. No agent got a heads-up. One HarperCollins staffer told WIRED there’s been no direction on what to say to authors or agents, because leadership seems to know it would start a firestorm.

If AI in a book is grounds for canceling the author, what’s AI in the back-cover copy of that book grounds for? If an author owes the reader a disclosure, why doesn’t the publisher?

Pick one rule and apply it to everybody.

My own view is that the tool doesn’t matter. The questions worth asking are whether the work is good, whether the facts hold and whether anyone was lied to about something they had a right to know. That standard works for authors and publishers alike. The current one only works for the people holding the contracts.

Can publishers put your manuscript into ChatGPT?

Authors should worry about this one most.

Every agent WIRED spoke with feared that publishers would paste authors’ manuscripts into public AI tools to churn out rejection letters or publicity descriptions. Some agents have started adding clauses to their contracts that forbid running the material through an AI system. The concern is plain. A public chatbot, as opposed to a closed system that stays inside the company’s own network, could end up with an author’s unpublished book.

There’s one encouraging detail. A HarperCollins source said several editors had asked whether they could paste a full manuscript into a public model, and the company’s lawyers told them no. Good. That should be written policy at every house, and it should be in the contract.

If you’re an author with a book under contract, ask. Ask your agent whether your contract has an AI clause. Ask your editor what tools touch your manuscript. You have every right to know where your unpublished work goes, and nobody’s going to volunteer it.

Why are publishers turning to AI?

Part of it is the people at the top, and part of it is the people they let go.

Staff across the Big Five told WIRED that layoffs gutted their teams, leaving skeleton crews of five or six people producing two or three times as many books. Some say they reach for AI only because there’s no other way to keep up. One put the blame where it belongs: not on the coworkers taking shortcuts, but on the company forcing them to.

Ownership matters too. Simon & Schuster has been owned by the investment firm KKR since 2023. Workers there heard the company was trialing Skan AI, software sold to leaders looking to cut costs and find work to automate, reportedly at KKR’s urging. Employees started an open letter to CEO Greg Greeley opposing it. Greeley’s memo in response said no decision had been made to use that product or any similar monitoring tool. WIRED adds that an AI detector flagged parts of his memo as machine-written.

Dan Sinykin, who wrote a book on how conglomerates reshaped publishing, told WIRED that publishing’s executives have more in common with other corporate executives than with their own editors, and that they’re testing what they can get away with before staff revolt. That fits a pattern I’ve traced through a lot of American business in The Enshittification of America. Cut the people, squeeze the ones left and hand them a tool to cover the gap.

Is AI in publishing the problem, or the secrecy?

The secrecy.

I’ve said it plenty of times on this site. It’s a tool. Use it or don’t. A publicist who drafts a pitch with AI and then rewrites it into something good has done her job. An editor who uses it to organize notes hasn’t betrayed anybody.

AI on its own produces flat, mediocre work, and a pitch letter full of the same tired constructions is bad whether a person or a machine wrote it. The people who notice the tells, like the agents comparing rejection letters, are judging quality, and quality is a fair thing to judge. One of those supposed tells is older than AI by a few centuries, as I showed in It’s Not X, It’s Y.

What doesn’t hold up is the double standard. Publishers can’t keep presenting themselves as the guardians of human-made books while running their own operations through the same software they cancel authors for. Agent Emma Dries told WIRED that what the industry needs most is for publishers to say clearly what’s being used and when. She’s right.

And if a publisher truly believes AI has no place near a book, it can say so and live by it. Some smaller presses are making exactly that their selling point. Meg Reid of Hub City Press told WIRED she expects the big houses to lean on AI for marketing, and smaller publishers to insist that the people are the point. That’s an honest position, and so is using the tools openly. Hiding them while punishing authors isn’t.

One rule for everybody

Writers aren’t asking for much. Tell authors what tools touch their books. Keep unpublished manuscripts out of public AI systems, in writing. And stop canceling authors for using a tool your own staff is using down the hall.

Until that happens, the safest thing an author can do is ask questions and read the contract. The AI and writing hub covers how to use these tools without losing your voice, and how to protect yourself when somebody else is using them on your work.

Frequently Asked Questions