.png)
AI in Publishing Is Becoming Infrastructure. Authors Need the Rules to Catch Up
AI is no longer sitting at the edge of publishing. It is moving into the plumbing.
That is the clearest signal from the latest publishing-industry reporting. Publishers are using AI for metadata, royalty statements, proofreading, customer service, forecasting, formatting and other back-office work. Publishers Weekly reports that 63% of industry professionals surveyed said their organisations were already using AI, while a separate forthcoming Book Industry Study Group and BookNet Canada report landed on the same figure.
The full Publishers Weekly feature is worth reading here: Publishing’s AI Reckoning.
The question has changed
For writers, this matters because the old argument whether AI should be allowed anywhere near books is already being overtaken by reality. AI is inside publishing workflows now. The more useful question is where the boundary sits between legitimate assistance and creative substitution.
That boundary is becoming visible. Publishers are increasingly comfortable with AI that helps organise, analyse, format or administer work. They are much less comfortable when it starts replacing the author, illustrator, editor or designer at the creative core of the book.
That distinction is close to my own position as an author. I write my books. I am comfortable using AI around the creative business of those books. And for research support, visualisation, marketing, organisation and administrative work it can all be useful. But that is not the same thing as handing authorship to a machine.
Trust is becoming part of the publishing workflow
The difficult part is that the middle ground is messy. Developmental assistance, rewriting, translation, editing suggestions and generative production tools can blur together very quickly. The Bookseller argued on 11 September that the debate is moving from whether AI belongs in publishing to where the line should be drawn, and that technology itself cannot reliably make that distinction for us.
Read the UK perspective here: AI and licensing: the debate is becoming a market.
That is important because detection is not a substitute for trust. Publishers have already faced controversies involving manuscripts suspected of AI use, while authors have challenged the accuracy and fairness of detection systems. A publishing industry built around suspicion would be a poor outcome for everyone particularly when false accusations can damage an author long before the technology is proven right or wrong.
The second shift: licensing is becoming a real market
The other important change is commercial. Licensing copyrighted material for AI is no longer just a theoretical alternative to lawsuits. Publishers are beginning to build actual licensing structures, and that changes the conversation for authors.
Publishers Weekly notes that HarperCollins has publicly licensed titles for AI training under an opt-in arrangement, while Wiley has reported tens of millions of dollars in AI-licensing revenue. Princeton University Press has also partnered with licensing platform Cashmere on a model designed around controlled, copyright-compliant use of books for AI research and retrieval.
For an example of that approach, see Princeton University Press Partners with Cashmere on AI Licensing.
This is where authors need to pay attention. If licensing grows, the argument will not only be about whether our work is used. It will be about consent, contract wording, payment, rights splits, rights reversions and whether authors are actually represented when publishers negotiate with AI companies.
What authors should watch now
Your contract language. AI clauses are becoming part of the rights conversation. Know what you have granted and what you have not.
The distinction between assistance and authorship. Using AI to organise research or help market a book is not the same as asking a model to write the novel.
Transparency without witch-hunts. Disclosure can be useful, but unreliable detection should not become a shortcut to accusation.
Licensing economics. If publishers make money from AI licensing, authors need clarity about how consent and remuneration work.
AI is becoming ordinary and that makes the rules more important
There is a paradox here. The more ordinary AI becomes inside publishing, the less useful the old all-or-nothing argument becomes. The industry is not choosing between an AI future and a non-AI future. It is choosing which uses it will normalise, which it will reject, and which it will license.
For authors, that is probably a better argument to be having. We do not need to pretend the technology will disappear. We do need to insist that efficiency is not allowed to erase authorship, that convenience is not confused with consent, and that the people who create the books remain part of the value chain built around them.
The real test is no longer whether AI belongs in publishing. It is whether publishing can use AI without losing the human judgement, trust and creative ownership that make books worth publishing in the first place.
Quick Shortcuts
The question for writers
If AI is already becoming normal infrastructure inside publishing, where would you draw the line between useful assistance and a machine doing work that should remain human?
About Rob Frankson
Rob Frankson is an author and the creator of 121 Minutes, writing about science fiction, creativity, publishing and the changing relationship between authors and artificial intelligence. About Rob
AI & Editorial Transparency
AI & The Author is edited and published by Rob Frankson. Artificial intelligence is used to assist with news research, initial drafting, content organisation and supporting imagery. All articles are reviewed and, where necessary, edited by Rob Frankson before publication. The opinions, editorial position and final decision to publish remain the author's.
Comments