Publishers use AI behind the scenes as staff push back, WIRED finds

AI is helping sell books at HarperCollins, Simon & Schuster, and Hachette—often without public disclosure or author consent, according to WIRED’s October 9 investigation. Workers describe tools writing publicity copy and emails to literary agents, while raising concerns about manuscript confidentiality, heavier workloads, and software they fear could help replace them.

Reporter Adam Morgan interviewed more than two dozen publishing workers, all anonymously to protect their employment. Their accounts describe executives encouraging AI adoption, employees struggling to establish boundaries, and resistance that has reached company town halls and a letter to Simon & Schuster’s chief executive.

Licenses first, use cases later

At HarperCollins, a current employee said senior leaders bought Claude licenses months ago, then struggled to implement them. Dozens of junior and senior employees were assigned to become “AI Champions,” identifying uses in monthly brainstorming sessions across the company. Staff also described purchases of ChatGPT and Jasper licenses.

An attendee said concerns about AI’s legal, ethical, and environmental effects were dismissed at a spring brainstorming session as “the cost of doing business.” One HarperCollins employee described resentment that the company had spent money on software workers did not want rather than addressing their longstanding requests for better pay.

Publicity copy is among the most common uses workers described across the three publishers. That writing appears on back covers, Amazon, and Goodreads, and helps pitch books to journalists, booksellers, and critics. A publicist interviewed by Morgan objected that AI-generated pitches undermine good-faith engagement with readers and media colleagues.

Simon & Schuster held an employee AI-use contest in May 2024, offering a $10,000 top prize and two $5,000 runner-up awards, staff said. One employee said the company later eased some of its AI push after repeated town-hall complaints. Other workers said layoffs and doubled or tripled book workloads had made AI shortcuts harder to resist.

A monitoring proposal sharpens the resistance

In recent weeks, Simon & Schuster employees began gathering signatures on a letter opposing Skan AI or similar software in company workflows. Beyond surveillance concerns, workers feared its automation-discovery tools could be used to shadow their work and identify opportunities to replace them.

On October 6, CEO Greg Greeley confirmed in a memo that the company had explored Skan AI’s Blueprint product. But no decision had been made to use it or another productivity-monitoring or workflow-analysis tool. Greeley also said he had not seen the exploration materials before concerns reached him the previous day.

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

Emma Dries, agent at Triangle House Literary, speaking to WIRED

Publishers draw different boundaries

Simon & Schuster’s marketing and communications chief, Wibke Grutjen, said employees have access to a limited set of enterprise AI tools. Use is optional, unapproved tools are prohibited, and the stated goal is more time for creative work. Hachette said it supports operational uses that help books reach readers, but not creative uses, including communications with authors and partners. HarperCollins did not respond.

Manuscript handling remains a particular concern for agents. One told WIRED that contract clauses barring material from being run through a language model apply to publishers as well as technology companies. A HarperCollins source said editors who asked to paste full manuscripts into externally connected AI tools were told not to by the legal team.

That account describes a restriction, not proof that unpublished manuscripts were submitted. The disclosure gap persists: a current HarperCollins employee said staff had received no direction on what to tell authors or agents about AI use.