At least three Big Five publishers quietly use Claude and ChatGPT, WIRED reports

While American book publishers have made headlines this year for suing Google over copyright and canceling books by authors suspected of using LLMs, a WIRED investigation says at least three of the Big Five US publishers, HarperCollins, Simon & Schuster and Hachette, have been quietly bringing AI tools into their work without public disclosure or author consent. The reporting rests on interviews with more than two dozen workers, all anonymous to protect their jobs. They say publishers use LLMs such as Claude and ChatGPT to email literary agents, write publicity and back-cover copy, design cover art and generate marketing materials, often at the direction of executives.

At HarperCollins, senior leaders bought Claude licenses from Anthropic months ago but, per a current employee, "very quickly realized they didn't know how to implement them." Dozens of junior and senior staff were then "voluntold" to become "AI Champions" and look for use cases in monthly brainstorming sessions. This spring, when employees raised legal, ethical and environmental concerns at one session, the person leading it dismissed them as "the cost of doing business," according to an attendee. Staffers say HarperCollins has also bought licenses for ChatGPT and Jasper, an agentic marketing platform. One division is "all in," including AI-generated marketing videos. One employee called the push "such a slap in the face," noting that money went to software instead of salaries; the average entry-level salary for New York City publishing employees at the Big Five and Scholastic was $47,583 in 2023. HarperCollins did not respond to requests for comment.

Across HarperCollins, Simon & Schuster and Hachette, one of the most common uses is publicity copy. A Big Five publicist told WIRED that using AI for it is short-sighted. Meg Reid, publisher of the independent Hub City Press, said LLM-written marketing for the Big Five "feels pretty inevitable." Literary agents also began suspecting this year that specific Big Five editors were using AI to write rejection letters, and employees at all three publishers say the agents were right. Simon & Schuster staff attended a workshop with OpenAI representatives, where one recommendation was composing emails to agents; the practice has also spread among editors at HarperCollins and Hachette. In May 2024 Simon & Schuster ran an employee contest for the best AI use cases, with $10,000 for the winner and $5,000 each for two runners-up, and has since rolled back some of the AI push after complaints in town halls, one employee says. Several staffers say AI assets are appearing in cover art, and some foreign presses distributed by Simon & Schuster openly acknowledge AI use in designs.

Company responses are limited. A Hachette spokesperson said the company supports AI for operational purposes but does "not support creative uses of AI, including communicating with our authors and partners." Wibke Grutjen, Simon & Schuster's global chief marketing and communications officer, said employees have access to a limited number of enterprise-level AI tools, use is not mandatory, and unvetted tools are prohibited.

The sharpest conflict is at Simon & Schuster. In recent weeks word spread that the company was running an experimental trial of Skan AI software, at the apparent urging of KKR, the investment firm that bought the publisher in 2023. Skan AI's own website says it gives leaders the evidence to reduce costs and prioritize automation. Employees began collecting signatures on an open letter to CEO Greg Greeley expressing "strong opposition" and asking for a commitment not to bring Skan AI or similar software into the company's workflows. Many were worried about surveillance and about "automation discovery" tools being used to shadow and replace them. New York magazine's Book Gossip newsletter reported the letter on October 6. The same day Greeley sent a memo confirming the company had been exploring Skan AI's Blueprint product but saying "no decision has been made to use Blueprint or any other productivity-monitoring or workflow-analysis tool," and that he had not seen the materials about the early-stage exploration until concerns were raised the day before. The AI-detection platform Pangram indicates at least some of the memo was written by AI.

Agents told WIRED they fear publishers may paste authors' manuscripts into LLMs to produce things like rejection letters and publicity descriptions, and some have added contract clauses saying material cannot be run through an LLM. One HarperCollins source says multiple editors asked managers whether they could paste a full manuscript into an open-loop LLM and were told not to by the legal team. Some staffers across the Big Five say they experiment with LLMs reluctantly because executives cut teams through layoffs and then tripled workloads; one employee described teams cut to five or six people, give or take, doing double or triple the number of books. Hachette employees expect AI policies to evolve under new chief information officer Carl Hixson, formerly chief technology officer at SRMG, a Saudi state-backed company that has embraced agentic AI. Editors also say they have seen LLM-translated manuscripts with AI-use disclosures at the Frankfurt and London book fairs, and HarperCollins CEO Brian Murray said at a UBS conference almost two years ago that AI presents "more opportunities than risks."

Dan Sinykin, author of Big Fiction, argues that Big Five executives "are trying to see what they can get away with before there's a revolt." Most industry people WIRED spoke with want clear AI usage guidelines and disclosures. Agent Emma Dries said, "We need publishers to be clear about what is being used and when."

Key facts

WIRED, citing more than two dozen anonymous workers, says at least three Big Five US publishers (HarperCollins, Simon & Schuster, Hachette) quietly use AI tools without public disclosure or author consent.
Reported uses include LLMs such as Claude and ChatGPT for emails to literary agents, publicity and back-cover copy, cover art and marketing materials; HarperCollins staffers say it also bought ChatGPT and Jasper licenses.
Simon & Schuster employees circulated an open letter against a Skan AI trial; CEO Greg Greeley's October 6 memo said no decision has been made to use Blueprint or any similar tool.
Staff cite layoffs and heavier workloads as the reason for reluctant experimentation, and literary agents are adding contract clauses barring manuscripts from being run through LLMs.
Hachette says it does not support creative uses of AI; Simon & Schuster says its tools are limited, vetted and not mandatory; HarperCollins did not respond.

Why it matters

Publishers have publicly appeared to resist generative AI, through the Google copyright suit and canceled books, while the WIRED reporting describes the opposite practice inside the Big Five: tools used in daily work without telling authors or the public. The gap between the public stance and the internal practice is the story. It also shows how AI adoption is arriving in a creative industry: pushed from the top, with staff asked to find use cases, in teams that were already cut and given heavier loads.

Who it affects

Authors and literary agents, whose manuscripts and correspondence may pass through tools they were never told about; agents already worry that rejection letters and publicity descriptions are written with LLMs. Editors, publicists, designers and translators at the publishers, who describe being pushed to adopt AI and fear being shadowed and replaced by automation tools. Independent presses and readers also sit in the picture, since Reid describes a split between corporate publishers seeking efficiencies and smaller ones insisting the humans are the point.

How to use it

This is a reported investigation, not a product, so there is nothing to adopt. What the source offers is practical context. Agents quoted say they have added contract clauses stating that material cannot be run through an LLM, noting that this applies to the publisher as well as to tech companies. The source also explains the difference between open-loop LLMs, which share data with the AI company's cloud, and closed-loop models, which run on an air-gapped computer or network. The industry professionals WIRED spoke with mostly ask for clear AI usage guidelines and disclosures.

How solid is it

The reporting is based on interviews with more than two dozen workers across publishers, which gives it breadth, but all of them are anonymous, and the internal materials it cites, such as Greeley's memo and the employees' open letter, are quoted only in part. Some details are second-hand: the Skan AI trial and KKR's role are described as word among Simon & Schuster workers and an apparent urging, and the contest and cover art claims come from staffers. Greeley's memo is described by WIRED, and the Pangram result only indicates that some of it was written by AI. HarperCollins, where much of the detail originates, did not respond; Hachette and Simon & Schuster gave statements that describe limited, vetted or operational use.

Risks and caveats

The source does not say any author's manuscript was actually fed into an LLM; agents express fears, and HarperCollins legal told editors not to paste manuscripts into open-loop tools. It does not say Simon & Schuster adopted Skan AI, only that Greeley's memo says no decision has been made. No count is given of canceled books or of books with AI-generated cover art or copy. Hachette's expected policy change under Carl Hixson is staff expectation, not a stated plan. Quotes such as Sinykin's, that executives are testing what they can get away with, are opinion rather than established fact.

“We need publishers to be clear about what is being used and when”

— Emma Dries, an agent at Triangle House Literary who previously worked in editorial at HarperCollins and Penguin Random House