Straight answer: nobody, including Amazon, has ever published an exact count of AI written books on the site, but every credible estimate puts the number in the hundreds of thousands and rising fast, out of a Kindle store that holds somewhere north of seven million titles. What matters more than the exact figure is what it’s doing to discovery, trust, and the pay for the human authors underneath it.
Nobody has the real number, and that’s the problem
I want to get this out of the way early because most posts on this topic quote a single headline number and treat it as gospel. There isn’t one. Amazon does not publish a public dashboard showing “AI generated titles: X”. What we have instead is a patchwork of reported studies, disclosure filings, and researcher scraping projects, each measuring something slightly different.
In February 2023, Reuters reported more than 200 e-books in the Kindle store listed ChatGPT as an author or co-author. That’s the number most journalists still quote. It was true at the time, and it was already a wild undercount even then, because it only counted books where someone had voluntarily credited the tool. Nobody was checking the other tens of thousands of low-content journals, romance novellas, and “how to” guides that used AI without saying so.
By September 2023, Amazon had seen enough spam that it changed its Kindle Direct Publishing rules: authors are now required to disclose AI-generated content when they submit a new book, and Amazon capped self-publishers to three new title uploads a day per account, specifically because bulk AI-generated dumping had become a problem worth writing a policy for. You don’t write a rule like that unless the volume behind it is large.
The numbers we do have, and what they add up to
Pull the threads together and this is the honest picture as of 2026:
- The Kindle store holds an estimated 7 to 10 million titles, the large majority of them self-published through KDP rather than traditionally published.
- Independent researchers scraping Amazon’s low-content and non-fiction categories in 2023 and 2024 found AI-generated books making up somewhere between 5% and 15% of new releases in specific categories such as children’s picture books, journals, and “get rich” style guides, far higher than in literary fiction.
- The Alliance of Independent Authors, the trade body that watches this space closely, has repeatedly flagged mass-produced AI books as a “bad actor” problem specifically in low-content and quick-turnaround niches, not across the whole store.
- Amazon’s own three-books-a-day cap tells you the scale of the bulk-upload problem was large enough to need a hard technical limit, not a polite request.
So the honest range is: hundreds of thousands of AI written or AI-assisted titles already live on Amazon, concentrated heavily in a handful of genres, and growing every month. Anyone giving you a single precise figure is rounding a moving target to make a headline work.
What happened when I tried this myself
Two years into rebuilding my own business in public, I ran a small experiment because I wanted to understand the mechanics rather than the theory. I drafted a short business guide, used an AI tool to help structure and speed up the first pass, then rewrote every section in my own voice with my own examples and numbers, the way I do with client work. I published it on Kindle as a straightforward test.
Two things surprised me. First, Amazon’s algorithm did not care one bit whether AI had touched the first draft. What it cared about was the same thing it always has: category, keywords in the title and back-of-book metadata, and whether people who clicked bought and kept reading. Getting the category and keyword search terms right mattered ten times more to my sales rank than the drafting method behind the words.
Second, and this is the part that stung a little, when I looked at the “customers also bought” carousel next to my own book, at least four of the six titles sitting alongside mine were clearly mass-produced AI content: same cover template, same generic subtitle formula, same twelve-chapter structure that reads like it was generated from a prompt asking for a twelve-chapter structure. They weren’t bad enough to get flagged, and they weren’t good enough to hold a reader past chapter two, but they were selling, because they were priced at 99p and stacked in Kindle Unlimited where the whole game is page reads, not book sales.
Who is flooding the store
It is not, mostly, working novelists trying to cut corners. The volume comes from a different crowd entirely: people chasing a side income who were sold the idea that publishing an AI book is a route to fast money with almost no effort. It is the same audience that gets pitched dropshipping courses and print-on-demand mugs, the same energy you’ll recognise from any list of realistic passive income ideas for young adults that separates what pays from what wastes a year. Publish-a-book-a-day schemes get sold hard in exactly the same Facebook groups that promise stay-at-home mom jobs that pay, and for the same reason: it sounds achievable, it costs nothing to start, and the platform doesn’t stop you from trying.
The uncomfortable bit, the part most articles about this dance around, is that a meaningful chunk of these low-quality AI books are not failing. They’re not competing against your literary novel. They’re competing against each other for a customer who never planned to read the whole thing anyway, someone who wanted a 40-page journal prompt book or a “learn Excel in a weekend” guide for £2.99 and got exactly the level of quality they paid for. Pretending every AI book on Amazon is a scam that fools nobody lets human authors off the hook for the actual threat, which is quieter: category flooding, keyword squatting, and reviews getting diluted by volume, not deception.
The uncomfortable truth about quality and discovery
Here’s the part that annoys careful writers most: in several genres, readers cannot tell the difference between a competent AI-assisted book and a mediocre human one, and a large share of them don’t care enough to check. Cozy mystery, short romance, business how-to, children’s colouring books, these are volume categories built on price and cover appeal, not on prose that rewards close reading. If your book lives in one of those categories and you were relying on “my writing is better” as your entire strategy, that strategy was already thin before AI turned up. It’s now visibly not enough.
Where craft still wins outright is where readers are buying a voice, not information: literary fiction, memoir, niche non-fiction from someone with real authority in the subject, anything where the author’s name is the product. That’s exactly the ground authors need to defend now, and it’s the same ground I talk about when I explain getting found and getting clients through a real digital strategy rather than through algorithm luck. A recognisable name, an actual email list, a body of interviews, quotes, and appearances that Google and readers can both trace back to you, is worth more defence right now than a slightly better first chapter.
What this means for working authors
If you self-publish on Amazon, here’s what I’d change starting this month:
- Stop competing purely on volume in low-content categories. If your business model depends on churning out short journals or generic how-to guides faster than the AI flood, you have already lost that race on cost. Move up a tier into books that need a real, verifiable point of view.
- Put your name and your evidence everywhere. Author bio, “look inside” sample, back matter, all of it should make it obvious a specific human with specific experience wrote this. Amazon’s disclosure rule works in your favour here if you use it: opt to clearly state your book is human-written where true, because that’s now a genuine trust signal, not a throwaway line.
- Check your competitor set before you launch. Spend an hour reading the “also bought” carousel in your category. If four of six competitors are template-driven AI churn priced at 99p, you are not pricing or positioning against them, you are pricing against readers who want more than that and are currently underserved.
- Treat your book income like a small business, because it is one. Royalty statements, ad spend on Amazon Ads, cover design costs, all of it needs proper records, the same way I tell every client to keep receipts and understand what you really need to keep for your tax return rather than guessing at year end.
- Watch for impersonation, not just competition. Publishing expert Jane Friedman found AI-generated books fraudulently published under her own name on Amazon in 2023, ranking on Google before her real work did. If you have any public profile at all, search your own name on Amazon every few months. It costs nothing and takes five minutes.
How to spot an AI written book before you buy or benchmark against one
You won’t get certainty, but these five checks get you close in under two minutes:
- Read the “look inside” sample past the first page. Generic AI prose tends to over-explain simple ideas and under-deliver on specific detail, real names, real numbers, real dates.
- Check the author page. A blank bio, no photo, no other work, no social links, is a strong tell.
- Look at publishing velocity. An author with fifteen books released in the same three months across unrelated topics is running a content operation, not writing books.
- Scan the reviews for pattern language, five-star reviews that read like they were also generated, all praising the same vague qualities with no specific chapter or example mentioned.
- Check the price and category together. Sub-£1 pricing stacked deep in a crowded Kindle Unlimited category is where the volume plays live, almost by definition.
Where this leaves authors who do it
I’m not anti-AI in publishing. I use these tools myself for research and structure, the same way I’d recommend any small business owner learn to use them well rather than pretend they don’t exist, which is why I point clients toward proper AI help for small business rather than either extreme of banning it or handing the whole job over to it. The line that matters isn’t “did AI touch this book anywhere.” It’s “does a real, checkable human stand behind what’s in it, with something to lose if it’s wrong.” That distinction is what readers, and increasingly Amazon’s own policy, are starting to reward. Volume was never a moat for authors. It’s an even worse one now.
Frequently asked questions
How many AI written books are on Amazon right now?
There is no official public count. Reuters found over 200 books listing ChatGPT as author or co-author back in February 2023, but independent researchers scanning low-content and quick-guide categories put the true figure, including undisclosed AI content, in the hundreds of thousands out of roughly 7 to 10 million total Kindle titles.
Does Amazon require authors to disclose AI-generated content?
Yes. Since September 2023, Kindle Direct Publishing has required authors to declare whether a book contains AI-generated text, images, or translation when they submit it, and Amazon capped new title uploads to three per day per account to curb bulk AI publishing.
Can readers tell if a book was written by AI?
In genres built on price and volume, such as short romance, low-content journals, and generic how-to guides, most readers cannot tell and many don’t check. In memoir, literary fiction, and expert non-fiction, where the author’s specific experience is the product, gaps in detail and authority are much easier to spot.
Should self-published authors stop using AI tools altogether?
No, but the win isn’t in using AI to write faster, it’s in using it for structure and research while keeping your own voice, evidence, and name clearly attached to the final book, since that authorship is exactly what’s becoming the scarce, valuable thing on Amazon.
