Amazon KDP and the AI Disclosure Requirement

If you self-publish on Amazon KDP and use artificial intelligence anywhere in your book, you've had one clear obligation since September 2023: disclose it. Not publicly, not on your book's product page, only to Amazon, through a checkbox in the publishing flow. But the line between what counts as AI content and what doesn't is less obvious than it sounds, and since 2025 enforcement has gotten a lot stricter than it was at launch.

If you used AI for brainstorming, a quick summary, or a grammar check, you probably don't need to disclose anything. If AI generated the actual text of a chapter, even if you heavily rewrote it afterward, you probably do. The whole distinction comes down to one detail that a lot of authors still get wrong.

What Does the KDP Policy Actually Say?

The core rule splits AI use into two categories, and which one your content falls into determines whether you check the box.

AI-generated content (disclose it). This is text, images, or translations that an AI tool produced from your prompt. It still counts as AI-generated even if you substantially edited it afterward: what matters isn't how much you changed it later, it's who created the first version (KDP Builder).

AI-assisted content (don't disclose it). This is text you wrote yourself, where AI played a supporting role: brainstorming ideas, drafting an outline, checking grammar, suggesting edits on something you already wrote. If the final content is yours and AI only helped polish it, the requirement doesn't kick in (Amazon Sellers Attorney).

You'll find the checkbox on the Title Details page of the publishing flow, in the section dedicated to AI content, and it shows up both for new titles and when you edit an already-published one. According to several independent guides that report the same wording, the text reads roughly: "This title contains AI-generated content. This includes text (for example, chapters, sections), images (for example, illustrations, cover art), or translations produced with AI tools."

If you're wondering exactly where legitimate AI help ends and AI actually creating content in your place begins, that's the same question at the center of Using AI to Write a Novel: Where It Helps (and Where It Doesn't), another article on our blog.

Why Did Amazon Introduce This Rule?

The policy didn't come out of nowhere. It followed months of negotiation with The Authors Guild, the most influential writers' organization in the US. When the policy was announced, the Guild called it "a welcome first step," explaining it was meant to protect readers from unknowingly buying AI-generated books and to protect human authors from being drowned out by a wave of cheap, mass-produced titles (The Authors Guild).

It wasn't an isolated move. The same year the policy launched, there was public discussion of a wave of low-quality ebooks, mass-generated and uploaded at a pace no human author could match, that was starting to clog up parts of the Kindle catalog (GeekWire).

What Happens If You Don't Disclose?

Things have changed a lot between 2023 and now. In the early days, the policy was little more than an honor-system request. By 2025 and 2026, enforcement got significantly stricter: Amazon's systems now analyze writing style, metadata, and upload speed to flag undisclosed AI books, with human review on flagged titles. The escalation path described consistently across industry guides looks like this: a first notice asking you to update your disclosure, suspension of the title during review, permanent removal if the violation is confirmed, a flag on your account that stays visible for future publications, and, for repeated or severe cases, suspension of the entire account (KDP Builder).

One detail that catches a lot of authors with an existing catalog off guard: review isn't limited to new titles. Amazon has started auditing books published before disclosure became mandatory, flagging ones that look AI-generated even if they came out years earlier (KDP Builder). If you have a backlist, this isn't just a problem for new releases.

At the same time, to push back against mass-publishing operations uploading dozens of titles a day, Amazon introduced a limit of 3 new titles per day per account, regardless of whether they contain AI content or not: a measure aimed at slowing down bulk producers, not authors publishing at a normal pace (KDP Builder).

Failing to disclose content that should have been disclosed is a violation of the terms of service, even if the content itself would have been fine with the correct disclosure. In other words, AI isn't banned. Hiding it is.

One last scope note: this policy covers books, ebooks and print, published on KDP. Audiobooks distributed through ACX follow their own separate guidelines.

Is the Disclosure Visible to Readers?

No, not on the product page. Amazon has stated the information is used internally, for its own editorial and policy decisions, and doesn't appear as a public label on the book's listing or as a badge readers can see. Amazon has also clarified that the disclosure itself doesn't affect royalties or search ranking: your book competes on the same criteria as always, quality, reviews, and relevance (KDP Builder).

The Authors Guild sees this as a limitation of the current policy, not a feature. The organization keeps pushing for the information to become visible to readers too, so they can make an informed choice, not just something Amazon uses internally. So far, Amazon hasn't agreed to that.

What This Means in Practice, If You Self-Publish

A few concrete pointers, so you don't have to reread the full policy every time:

• Brainstorming, outlining, preliminary research: AI-assisted, no disclosure needed, as long as you write the final text yourself.
• Grammar fixes or editing suggestions on text you already wrote: AI-assisted, no disclosure needed.
• A chapter, scene, or paragraph AI generated from a prompt, even if you heavily rewrite it afterward: AI-generated, disclosure required.
• Cover art or interior illustrations generated with an AI tool: AI-generated, disclosure required.
• A translation of your book done by an AI tool: AI-generated, disclosure required.
• Not sure which category something falls into? The safest guideline, repeated across every source we checked, is to disclose anyway. The risk of over-disclosing is basically zero. The risk of skipping it, if it's discovered, ranges from a suspended title to a suspended account.

It's also worth keeping track of your writing process, even just saving intermediate drafts of your manuscript. Not only for the KDP disclosure: the same kind of documentation is what you need to establish who actually owns the rights to text written with AI's help, a topic we covered in depth in Who Owns AI-Generated Content?

Frequently Asked Questions

If I only use AI to check my grammar, do I need to disclose it?

No. That counts as AI-assisted: the text stays yours, AI just checked or corrected what you'd already written.

Does the disclosure show up on my book's page, visible to readers?

No. Amazon uses it internally for its own policy decisions, it doesn't appear on the product listing. The Authors Guild keeps asking for it to become visible to readers too, but for now it isn't.

Does the 3-titles-a-day limit apply even if I don't use AI?

Yes, it's a general account limit introduced to slow down mass publishing, it isn't tied to the AI disclosure itself. For an author publishing at a normal writing and editing pace, it's practically never an issue.

What happens if I disclose something as AI-generated out of caution, but I actually wrote it myself?

Amazon has stated the disclosure doesn't affect royalties or ranking, so disclosing out of caution, when you're unsure, doesn't create any real downside. It's the safer choice whenever you're not certain how to classify something.