What are Amazon KDP's AI content disclosure rules and how do I comply?

Amazon KDP's AI content disclosure rules require authors to tell Amazon, during the publishing process, whether their book contains AI-generated text, images, or translations. The policy, introduced in September 2023, splits AI involvement into two categories: AI-generated content, which must be disclosed, and AI-assisted content, which does not need to be disclosed. If you skip the disclosure questions or answer them dishonestly, you risk account termination. Below is a full breakdown of what the rules say, how to answer the questions correctly, and where the policy falls short.

The Direct Answer: What Amazon Requires

Also worth reading: What are the mandatory AI disclosure requirements for publishers and content creators in 2026? · What's the difference between AI-assisted and AI-generated content for Amazon KDP, and which should self-publishers use in 2026? · What are AI disclosure clauses in publishing contracts and how should authors handle them in 2026?

When you create a new Kindle eBook or paperback through KDP, Amazon asks a mandatory question in the publishing workflow: did you use AI tools in creating your book? You must answer for three specific elements: the text, the images, and the translations. If you select yes for any of them, you then specify whether the content was AI-generated or AI-assisted.

AI-generated content means text, images, or translations created by an AI-based tool, even if you edited it heavily afterward. This must be disclosed to Amazon. AI-assisted content means you created the work yourself but used AI tools to refine, edit, check grammar, or brainstorm. Amazon explicitly states that AI-assisted content does not need to be disclosed. The distinction matters enormously: a novel you wrote and then ran through Grammarly or ProWritingAid is AI-assisted and exempt. A novel drafted by ChatGPT and lightly edited by you is AI-generated and must be declared.

Amazon has stated that it asks about AI at the point of publishing rather than displaying AI labels to readers. As of 2026, Amazon has not rolled out a visible 'AI-generated' badge on Kindle product pages the way some retailers and stock-photo sites have. The disclosure goes to Amazon, not to the buying public, though Amazon reserves the right to use that information in enforcement and quality control.

Why Amazon Introduced the Rules

The policy arrived in September 2023 after KDP was flooded with AI-generated books, some of which were dangerous. The most cited example was a foraging guide, apparently AI-generated, that included inaccurate and potentially lethal advice about mushroom identification. Amazon capped Kindle eBook uploads at three titles per day in response to the volume problem, and the disclosure requirement followed shortly after.

Amazon's stated goals were twofold: protect readers from low-quality or harmful AI content, and protect authors whose work was being scraped to train AI models without consent. The second goal produced the policy's most concrete commitment: Amazon said it would inform authors when third parties request the right to use their work for AI training and give them the option to opt out. That commitment addressed a real grievance among writers, though critics noted it placed the burden on individual authors rather than on the platforms doing the scraping.

The broader context is a market under real strain. Industry reporting through 2024 and 2025 documented thousands of AI-generated titles appearing monthly across categories like low-content books, children's books, and niche nonfiction. Discoverability for human authors has degraded in saturated categories, and Amazon's enforcement has been inconsistent, which is why the disclosure rules alone have not solved the problem.

AI-Generated vs AI-Assisted: The Critical Distinction

This is where most authors get confused, so it deserves careful treatment. Amazon's definitions are narrower than many writers assume.

AI-generated means the AI produced the content itself. If ChatGPT, Claude, Gemini, Sudowrite, or any other generative tool wrote paragraphs, chapters, descriptions, or a full manuscript that you then published, that text is AI-generated regardless of how much you edited it. If Midjourney, DALL-E, or Stable Diffusion produced your cover art or interior illustrations, those images are AI-generated. If you used DeepL or an AI translation service to produce a foreign-language edition, that translation is AI-generated.

AI-assisted means the AI helped you but did not produce the content. Spelling and grammar checkers, research assistance, outlining brainstorming where you wrote every word, and formatting tools generally fall here. Amazon's own guidance says you created the work and used AI to refine it.

The gray zone is real and Amazon has not fully resolved it. If you dictate rough ideas and an AI restructures them into prose, is that generated or assisted? If you write a chapter and ask an AI to rewrite it in a different voice, most reasonable readings say generated. If you ask an AI to suggest ten plot directions and you write the book yourself, that is assisted. When in doubt, the safer legal and practical position is to disclose. There is no penalty for disclosing AI-generated content; there are penalties for failing to disclose it.

## How to Complete the Disclosure: Practical Steps

The disclosure happens during book setup in your KDP Bookshelf. When you create a new eBook or paperback, scroll to the AI content section in the publishing steps. Amazon asks: 'Did you use AI tools in creating texts, images, and/or translations in your book?' You answer yes or no for each of the three categories separately.

If you answer yes to any category, a follow-up question asks whether the content was AI-generated or AI-assisted. Select AI-generated if the tool produced the content. You do not need to name the specific tool, provide prompts, or quantify the percentage of AI involvement. The question is binary per category. For a book with a human-written manuscript and an AI-generated cover, you would answer no for text, yes for images, and specify the images as AI-generated.

For existing books published before the policy, Amazon has not required retroactive disclosure, and as of 2026 there is no mechanism demanding you go back and update old listings. However, if you republish or make significant changes to an older title, you will encounter the disclosure questions in the updated workflow. New editions, new formats, and new translations all trigger the questions fresh.

Keep your own records of how each book was produced. If Amazon ever questions a listing, being able to document your process, your drafts, and your tool usage is your best defense. Screenshots of your disclosure answers are worth keeping too, since KDP support sometimes has trouble locating what an author previously submitted.

Enforcement, Penalties, and Amazon's Limits

Amazon's stated penalty for failing to disclose AI-generated content is account-level action, up to and including termination of your KDP account. In practice, enforcement has been driven less by the disclosure questions and more by Amazon's separate content guidelines: books must be human-authored original work or properly disclosed, and mass-produced, low-quality AI spam can be blocked or removed regardless of disclosure status.

The enforcement reality is uneven. Amazon relies heavily on reader reports and automated detection, and neither catches most undisclosed AI books. Investigations by journalists and researchers have repeatedly found AI-generated children's books, health guides, and low-content journals on the platform with no disclosure. Amazon removes titles when flagged, but the volume problem outpaces removal. For honest authors, the practical risk is asymmetric: you face account termination for hiding AI use, while thousands of undisclosed AI books sit unpenalized because nobody reported them.

There is also a legal dimension beyond Amazon's rules. The US Copyright Office has stated that purely AI-generated material is not copyrightable, and its 2023 guidance and subsequent registration decisions require applicants to disclaim AI-generated portions of a work. A book that is substantially AI-generated may not be protectable at all, which means competitors could legally copy it. This is a risk Amazon's disclosure does not address and one authors frequently overlook.

Common Mistakes Authors Make

The most common mistake is assuming AI-assisted means 'I used AI at all.' Authors who used ChatGPT for brainstorming sometimes disclose as AI-generated when they did not need to, which is harmless but unnecessary. The opposite error is more dangerous: authors who had AI draft substantial portions of text and classify it as 'assisted' because they edited it. Editing does not convert generated content into assisted content under Amazon's definitions.

Another frequent error is forgetting images. Authors correctly disclose AI-written text but answer no for images because they think of their Midjourney cover as 'theirs' since they prompted it. Under Amazon's rules, AI-generated cover art must be disclosed. The same applies to AI-generated interior illustrations in children's books, which are among the most commonly flagged categories.

Authors also misunderstand translations. Using AI to translate your own book into Spanish or German makes that translation AI-generated, even though the underlying book is yours. If you hire a human translator who uses CAT tools with machine translation suggestions, the line blurs, but fully automated translation pipelines should be disclosed.

Finally, some authors believe disclosure will hurt their sales or visibility. There is no evidence Amazon demotes disclosed AI books in rankings, and the disclosure is not shown to readers. Hiding AI use, by contrast, carries termination risk and, if discovered publicly, reputational damage that has ended several self-publishing careers.

How KDP Compares to Other Platforms

Amazon is not alone in grappling with AI disclosure, and its approach is neither the strictest nor the loosest. Understanding the alternatives matters if you publish wide.

Draft2Digital distributes to multiple retailers and asks about AI-generated content during setup, with definitions broadly similar to Amazon's. Apple Books requires you to warrant that you have the rights to your content and has updated terms addressing AI-generated works, though its disclosure mechanics are less granular than KDP's. Google Play Books permits AI-generated content but requires disclosure and prohibits content that infringes or misleads. Smashwords, now part of Draft2Digital, has taken a harder line against mass-produced AI content. Stock image and audiobook platforms vary widely, with some banning AI voices entirely and others requiring labeling.

Amazon's choice to collect disclosure privately rather than label publicly has drawn criticism from both sides. Reader advocates argue buyers deserve to know; AI-friendly authors argue public labels would stigmatize legitimate tools. As of 2026, the private-disclosure model still holds, but regulatory pressure in the EU under the AI Act's transparency provisions may eventually force visible labeling across platforms.

When to Act and What It Costs

Compliance costs nothing. The disclosure questions are built into KDP's free publishing workflow and take under a minute per book. There is no fee, no verification process, and no certification to purchase. If you publish regularly, the only real cost is a few minutes of honest record-keeping per title.

Act now if you are publishing anything new, because the questions are mandatory and cannot be skipped. Act retroactively only if you are updating an existing book, launching a new format, or if you previously answered incorrectly and want to correct the record. Correcting a past answer is straightforward: contact KDP support or update the listing if the workflow allows it, and be honest about the original error. Voluntary correction is far safer than waiting for a reader report or an Amazon audit.

If you are using AI heavily in your workflow, the moment to formalize your process is before your next upload, not after. Document which tools touched which parts of the book, decide your generated-versus-assisted classifications in writing, and apply them consistently. Authors who treat disclosure as an afterthought are the ones who make mistakes under deadline pressure.

The Honest Assessment

Amazon's disclosure rules are a reasonable first step wrapped around a weak enforcement mechanism. The generated-versus-assisted distinction is sensible and mirrors how the Copyright Office thinks about authorship. Collecting disclosure at upload is low-friction. But the rules do not label anything for readers, do not cap the flood of AI content beyond the three-per-day upload limit, and depend on a reporting system that catches only a fraction of violations. The result is a policy that honest authors comply with and bad actors ignore, which is the classic weakness of honor-system regulation.

For you as an author, the calculus is still simple. Disclosure costs you nothing, protects your account, and takes a minute. The bigger strategic questions, whether AI-generated books are copyrightable, whether readers will tolerate undisclosed AI content, and whether regulators will force public labels, are still being decided. Publish with accurate disclosures now, keep records, and stay alert, because this policy will not look the same in two years as it does today.