Do professional authors use AI? What the numbers actually say
When BookBub asked 1,229 authors whether they use generative AI in their work, 45% said yes. That number bothered a lot of people. Is writing, as a hobby, a job and a craft, dying? Nearly half sounds like a lot.
Most of the coverage stopped at the split: about half for, about half against. But the rest of the survey is more interesting than the headline, and it tells a calmer story.
Writers aren’t special here
First, 45% isn’t an author thing. It’s a work thing.
Gallup has been asking American employees how often they use AI in their job since 2023. Around the time BookBub ran its survey, 40% of US workers said they used it at least a few times a year. By May 2026 it was 52%.
So authors aren’t being uniquely swamped by an army of AI agents. Work in general is folding AI in, and fast: Gallup’s number more than doubled in under three years. My bet is it ends up the way computers did. Nobody asks whether an author “uses a computer” anymore.
Does that mean you’ll stop writing yourself? Are tomorrow’s books all going to be written by AI? That’s where the details come in.
Most of it isn’t writing
BookBub also asked the authors who use AI what they use it for. Writing isn’t at the top. It’s sixth.
Research, marketing, plotting and editing come first. (I also suspect formatting, admin and paperwork would have ranked high if they’d been options.)
That’s the same list as the rest of the working world. Gallup’s employees put writing and editing (51%) and search or research (49%) at the top of theirs too.
Research is popular everywhere research is part of the job, because it’s a lot faster than digging through websites one tab at a time. It isn’t more or less trustworthy than a search engine, either. Both can give you wrong information, and both can be incomplete. If you google something and believe the first result, you’ll have a bad time too. Fake news existed long before AI. The rule is the same in both cases: check the source.
Writing with AI rarely means the whole book
Now look at the writing row more closely. Among authors who use AI, 30% never use it to write. Of those who do, most use it occasionally or rarely (41 points), not frequently (29).
Do the arithmetic across the whole survey, and about one author in eight uses AI to write frequently.
“Occasionally” doesn’t mean the AI writes the book. It’s a placeholder chapter so you can move on to the scene you actually want to write. It’s a description you’re stuck on, when the blank page wins the staring contest. The book is still the author’s.
The AI nobody counts
Then there’s the AI people use without thinking of it as AI.
A lot of authors use Canva, and especially its background remover, which is an AI tool. Plenty of them would answer “no” to “do you use AI for your marketing” in good faith. Then a platform flags their post as AI-made because of an edit, even though they paid an artist for the cover.
So technically they use AI for their marketing assets, the bulk of that material is still human-made art, and the survey can’t see any of it.
Where authors draw the line
So, do professional authors use AI in their work? More and more, yes. But some uses are far more divisive than others, and you can see it at the bottom of the chart: translation, audiobook narration and cover art have the most “never” answers, with writing close behind.
Drafting is the obvious one. Are you still writing a book if you’re prompting it? How much of the writing has to be yours for the book to feel like yours? There’s a legal side too: the US Copyright Office only protects what a human wrote, so the AI-written parts of a book may have no copyright at all.
Covers are about perception (what will readers think when they see one?) and about community. Cover artists are part of the author world, and a lot of writers feel conflicted about replacing them with a machine.
Translation has the same problem as drafting, and in a way it’s worse. You can’t even read the result to judge whether it still sounds like you.
Who does your marketing?
There’s one thing a survey like this can’t show. An author who says “yes” or “no” to AI for marketing is in a very different place depending on who does the marketing. If you’re self-published, it’s you. If you’re traditionally published, it’s mostly your publisher.
And if your publisher uses AI to sell your book, is that so different from using it yourself?
And publishers?
Publishers are about as divided as authors, and their uses cover the same range.
HarperCollins offered authors $2,500 per title to license some nonfiction backlist to an AI company for training, opt in only. Penguin Random House went the other way and added a line to its copyright pages forbidding AI training. In France, HarperCollins’s Harlequin started testing machine translation for some of its romance lines, with human translators reworking the output. The translators’ association called it a first at that scale.
Inside the houses, AI is already at work. This month Wired reported that staff at HarperCollins, Hachette and Simon & Schuster use ChatGPT and Claude for publicity copy, back-cover copy and marketing material. Hachette says it supports AI “for operational purposes but not creative uses”. As far as I can find, none of the Big Five has said it doesn’t use AI at all.
The novel itself is where they draw the line. The public cases so far went one way. In March, Hachette pulled Mia Ballard’s Shy Girl after readers raised suspicions and a New York Times investigation ran the text through AI detectors. The detection firm Pangram estimated most of it was machine-written. (Ballard denies writing it with AI and blames an editor she hired.) Macmillan has cancelled a book over suspected AI use too.
Which brings us to my favourite irony: Pangram is AI. The tool publishers reach for to keep AI out of books is itself AI.
Like authors, publishers can automate a lot, from proofreading to admin. And deciding not to when the competition does is a hard call for them too.
So how much AI belongs in your work?
Here’s the rule of thumb I use for anything I could automate:
- If it replaces the part where you bring the value, think hard. Is the speed worth what you’d lose?
- If it saves time and the result is the same as when you do it, it’s probably worth automating. Nobody’s reading your invoices for your voice.
- Watch for automation as a way out of people. AI can do a good job, but even at its best it usually isn’t as good as the best human. Sometimes we hand a task to a machine because delegating to a person is hard. People come with expectations, the odd conflict, the awkward email. But people also bring things a machine can’t: a perspective you don’t have, the editor who pushes back. Before you give the job to a machine, ask whether that friction is exactly what you need.
The authors in BookBub’s survey mostly landed in the same place. They use AI for the jobs around the book and keep the book for themselves. If that’s where you are too, I wrote about five ways to use AI if you like writing.