Amazon KDP's AI Disclosure Rule Now Covers Books You Already Published

You published your book two years ago. Back then, KDP didn't ask whether an AI tool drafted any of the prose, and the dashboard didn't have a box for it if you'd wanted to say so. In 2026, that question is no longer reserved for new uploads. Reports from across the self-publishing world describe already-published books getting pulled into content review, and authors finding out the hard way that their old answer, or their old silence, doesn't hold up anymore.

Amazon added AI disclosure to the KDP content guidelines back in 2023, aimed at new submissions. Since then, enforcement has widened well past that starting point. Multiple publishing industry guides report that Amazon reviews previously published titles as well, flagging books that read as AI-generated but were never marked that way. How Amazon finds those books isn't public. The guides point to a mix of automated detection and human review, with signals like writing style, metadata, and upload velocity, but that's inference from observed patterns, not anything Amazon has confirmed. That opacity is exactly why authors keep getting caught off guard.

The pattern being reported looks like this. A book that sold quietly for a year gets pulled for content review. Getting it reinstated means answering a disclosure question the author thought they'd already settled.

AI-Generated vs. AI-Assisted: Where Amazon Draws the Line

The actual rule is narrower than the anxiety around it suggests. Under Amazon KDP's policy, content is AI-generated if an AI tool created the text, images, or translation that appears in the published book, even if the author edited it afterward, and AI-assisted if the author created the content and used AI only to refine, check, or brainstorm it. Only AI-generated content requires disclosure.

An outline you built with a chatbot and then wrote yourself doesn't need disclosure. Grammar checking, line editing your own prose, brainstorming titles, none of that counts either. But a chapter an AI tool drafted, even one you heavily revised afterward, is still AI-generated under Amazon's definition, and needs to be disclosed. The same goes for cover art or interior illustrations made with an AI image tool, and for any translation an AI produced. Editing doesn't reclassify the work. Polishing word choice on an AI draft doesn't make it yours. Rewriting from scratch in your own voice, so that none of the AI's actual sentences survive in the book, is a different situation, since at that point you created the content that appears on the page. Amazon doesn't spell out where that line sits, so when your process lands anywhere near it, disclosure is the safer call.

Where this trips people up in practice is usually nonfiction. The clear cases sort themselves: a novelist who ran a full manuscript through an AI tool knows to disclose, and an author who wrote every sentence after using AI to organize research notes is genuinely in the clear. The murky case sits in between: someone who fed an AI a chapter's worth of bullet points, asked for prose, and tightened the language afterward. That's the scenario Amazon's rule was written for, and it's the one worth checking honestly rather than assuming your edits were enough.

How to Add an AI Disclosure to a Book You've Already Published

The update itself takes minutes. New titles see the disclosure question on the Content Details page during setup, worded roughly as whether the title contains AI-generated content, with a follow-up asking which category (text, images, or translation) applies.

For a title you already published, open the book on your Bookshelf, click into edit book details, and the same checkbox sits on the same page. One step catches people here: checking the box isn't enough on its own. You need to click through to the final Pricing page and hit Publish to submit the change, or the update sits in draft and nothing reaches Amazon. Once submitted, Amazon confirms the update within a day or two. If you don't see the disclosure field, usually because the title predates the 2023 rollout, contact KDP Support for guidance.

What matters most for anyone re-checking old titles is that this disclosure goes to Amazon, not to your readers. It doesn't appear on your product page, and Amazon has stated plainly that it doesn't affect royalties, search ranking, or category placement. Whether you tell your readers is a separate, voluntary call; the Authors Guild's AI best practices recommend a note in the front matter when AI generated part of a book, but no platform requires it. Checking the box honestly costs you nothing. Not checking it, when it turns out you should have, costs considerably more: undisclosed AI-generated content violates KDP's terms, Amazon can remove books over it, and repeated or serious violations can lead to account-level action. Authors have also reported withheld royalties on removed titles and difficulty republishing a book once it's been pulled.

Audit Your Back Catalog Before Amazon Does

If you've published anything in the last two or three years and used AI tools anywhere in your workflow, the practical move is to get ahead of the review rather than wait for it. Go through your Bookshelf title by title and decide honestly, section by section if you have to, what the AI actually produced versus what it only helped you produce. Update the disclosure on the ones that need it. And keep in mind that the answer you settle on here does double duty: it should match what you'd tell the U.S. Copyright Office, for reasons covered below.

It's also worth keeping some trail of your own process. Draft versions, outline notes, revision history. If a title does get pulled for review, you're pointing to evidence of your involvement rather than reconstructing your memory of a manuscript you wrote eighteen months ago under a support ticket deadline.

Disclosure and Copyright Are Two Different Questions

What you tell Amazon determines whether your book stays listed for sale. What the U.S. Copyright Office will actually register determines what you legally own in that manuscript, and those two things don't move together. The Office's registration guidance requires applicants to disclose more than de minimis AI-generated material, and the Authors Guild warns that knowingly hiding it on a registration application can amount to fraud on the Copyright Office. So the honest answer you give Amazon should match the honest answer you give the government. Disclosing to Amazon doesn't shrink your copyright, and staying quiet doesn't expand it. We've gone through the copyright side of AI-generated content in more detail separately, including where the line sits on what you can and can't register when AI wrote part of the work. If you're not sure what portion of a heavily AI-assisted manuscript is actually yours to protect, that's the piece worth reading next.

Quick Answers for Authors

Do you have to disclose using ChatGPT or Claude to edit your own writing? No, editing your own words is AI-assisted, not AI-generated.

Does the AI disclosure show on your Amazon product page? No, it stays between you and Amazon.

Does disclosing AI-generated content hurt sales or ranking? Amazon says it does not. The risk sits entirely with undisclosed AI-generated content, not with disclosure itself.

Amazon's guidance is still moving, and what counts as substantially rewritten today may get more specific tomorrow, so check KDP's content guidelines every few months rather than assuming last year's understanding still holds. When a title sits anywhere close to the gray area, the safer answer costs nothing. Check the box.