The Publishing Industry is Secretly Using Lots of AI
A recent industry survey noted AI adoption rate inside publishing is less than six percent, the lowest of any creative industry measured, according to coverage of a UK survey from Creative Access. Copyright and authorship, the report says, are what's holding the industry back.
A different survey tells a messier story. The Book Industry Study Group polled the wider book trade, publishers, libraries, retailers, service providers, and found nearly half already using AI tools somewhere in their operations. The same survey found a third of organizations refuse to touch it, permanently, on principle. Both numbers describe the same industry in the same year. The six percent and the near-half describe different things: one counts disclosed, official, customer-facing use; the other counts everything, including the behind-the-scenes jobs that never make it into a press release.
Academic publishing already shows what that gap looks like. On company runs an in-house AI system that screens submissions for suitability and conflicts of interest before an editor ever opens the file. A couple of academic presses, unnamed publicly, use AI to help identify peer reviewers. As one industry voice put it in a recent report on the topic: we're all using it, so just be open about it. None of that discrepancy appears in a survey asking whether AI writes anyone's sentences. It was never going to. The question and the reality are answering two different problems.
It’s ebooks all over again
This industry has made a version of this bet before, and I watched it from inside acquisitions. Publishing was slow on ebooks too: cautious about pricing, cautious about DRM, cautious about releasing a channel it had controlled for a century. The caution didn't buy what publishers hoped it would. Amazon built the channel on its own terms, while publishers were still debating whether a $9.99 ebook would destroy the paperback business. An entire generation of self-published authors built careers on that gap in the industry. Slow ebook adoption cost publishing real ground.
The AI question looks different. The ebook fight was about a channel: who controls distribution, who sets the price, who owns the relationship with the reader. Publishing lost pieces of all three and survived anyway, because a channel is infrastructure, and infrastructure can be rebuilt or rented.
The six percent is about authorship
The AI fight, for at least one piece of it, authorship, who physically wrote the sentences, who gets credited, who gets paid when a model can produce a passable draft, isn't infrastructure. It's the product. Losing ground on a channel is a business problem. Losing control of controling the rights to who wrote the book is a big IP problem.
Screening submissions, flagging conflicts of interest, cleaning up comp data: none of that is authorship, and publishing has been slower than it should have been on all of it, the same slowness that cost ground on ebooks. Whether an AI model gets credit for a sentence a reader thinks came from a person is a different question, and the industry has been right to treat it like one. My complaint is that publishing has applied one level of caution to two problems that don't deserve the same weight.