At least three of the five biggest US book publishers, HarperCollins, Simon & Schuster and Hachette, have been bringing AI tools into their work without telling the public or authors, according to WIRED's interviews with more than two dozen staff, all anonymous for fear of their jobs. They say executives are directing the use of models such as Claude and ChatGPT to write emails to literary agents, publicity and back-cover copy, cover art and marketing material. This sits oddly beside a year in which the same industry has been seen to fight AI, suing Google and cancelling books by authors suspected of using it.
HarperCollins staff describe licences bought from Anthropic before the company knew how to use them, followed by dozens of employees being named "AI Champions" and asked to find uses in regular brainstorming sessions. When staff raised legal, ethical and environmental worries, one was reportedly brushed off as the cost of doing business, and the staff portal later added a note playing down AI's environmental footprint. The company also holds ChatGPT and Jasper licences, and one division is described as fully committed, including AI-made marketing videos. At Simon & Schuster, workers sat through an OpenAI workshop and ran an internal contest in May 2024 with a $10,000 first prize, then rolled back some of the push after complaints at town halls. AI assets are also said to be appearing in cover art.
“We need publishers to be clear about what is being used and when”
Emma Dries, agent at Triangle House Literary, formerly in editorial at HarperCollins and Penguin Random House, to WIRED · wired.com ↗
Agents suspect that editors at the big houses have used AI for rejection letters, and staff at all three companies say they think those suspicions are right. Some agents have added contract clauses stating that material cannot be run through an LLM, worried that unpublished manuscripts could be fed to systems that are not closed-loop. At HarperCollins, editors who asked to paste a whole manuscript into an open model were told no by the legal team. Editors also say that accusations of AI use around books such as Hachette's Shy Girl and Macmillan's Call Me, I'll Hide the Body are only the visible part, with other cancelled books never reaching the news.
The salary backdrop matters to them: the average entry-level pay at the Big Five and Scholastic in New York was $47,583 in 2023, and one HarperCollins employee called the AI spending a slap in the face after years of arguing for decent wages. The OpenAI workshop at Simon & Schuster left an attendee feeling the presenter had no idea what a publisher does, and one of its suggestions for the book trade was writing emails to literary agents, a habit that has since spread at HarperCollins and Hachette too. Simon & Schuster's communications chief says the aim is to free staff for deeper creative work alongside authors.
Staff say the experimenting is partly forced: layoffs left teams of five or six handling double or triple the number of books. At Simon & Schuster the latest flashpoint is a trial of Skan AI, workflow-monitoring software reportedly encouraged by owner KKR, which prompted an open letter to CEO Greg Greeley opposing it and fearing it would be used to find work to automate. Greeley's memo of 6 October said no decision had been made. Hachette says it backs AI for operational uses but not creative ones, such as communicating with authors, and Simon & Schuster says its tools are optional and must be vetted. HarperCollins did not respond.
Dan Sinykin, a scholar of publishing consolidation, argues that the bosses of the big houses share more with executives at other large American companies than with the editors, publicists and translators who work for them, and are testing what they can get away with before staff revolt. Hub City Press's publisher expects AI-written marketing to feel inevitable at the Big Five, as corporate houses chase efficiencies and smaller presses insist that human involvement is the point. A Big Five publicist, for their part, calls using AI for publicity copy short-sighted, since publicists owe good faith to readers and to journalists who will have little patience for machine-written pitches.
There is a history of executive enthusiasm. HarperCollins chief executive Brian Murray said at a UBS conference almost two years ago that AI offered more opportunities than risks, and he expected it to help publish more translations and audiobooks. In August the executive publisher of independent house Europa Editions told booksellers in Baltimore he was excited about AI in translation. Staff at Hachette expect its policies to change under new chief information officer Carl Hixson, formerly technology chief at a Saudi state-backed company that has embraced agentic AI. WIRED adds that the detection tool Pangram flags part of Greeley's memo as AI-written.
Beyond the three companies, editors report seeing AI-translated manuscripts carrying disclosures at the Frankfurt and London book fairs. Most of those WIRED spoke to want clear guidelines and disclosure above all, and the Simon & Schuster letter is still collecting signatures. One HarperCollins staffer says there is no guidance on what to tell authors or agents, because leadership expects an uproar from them.
Sources · 1