At least three large US book publishers are using generative AI in publicity, marketing and design even as the industry challenges unauthorized model training and AI-assisted manuscripts. More than two dozen workers described the internal practices to WIRED on condition of anonymity.
Employees at HarperCollins, Simon & Schuster and Hachette reported uses that include emails to literary agents, back-cover text, publicity pitches, cover images and marketing videos. Some said executives directed junior and senior staff to find applications for newly purchased Claude, ChatGPT or Jasper licenses.
The conflict is partly about disclosure. Authors and readers may encounter AI-assisted packaging without being told, while staff say consent is not consistently requested. Publicists also worry that automated pitches weaken trust with booksellers, reviewers and journalists. Workers connected the spending to longstanding concerns about low entry-level pay and job quality.
The reporting does not establish that every imprint uses these tools or that AI produces final materials without human editing. The named companies have different policies and practices, and many accounts are anonymous because employees feared workplace consequences. Still, the examples show that adoption is moving beyond experimental chat use into the visible commercial work surrounding books, even where publishers publicly maintain stricter standards for authors.