Every book submitted to Amazon KDP now runs through a question most authors skip past without thinking: did AI generate any of this content? The honest answer, for most non-fiction authors using any modern writing tool, is "sort of" — and "sort of" is exactly the answer Amazon's policy was built to force into a real decision. Get the category wrong and you're not just filling out a form incorrectly; you're risking a blocked listing or a suspended account over a distinction most authors have never had explained to them clearly.
This isn't a ban on AI. Amazon has been explicit that using AI tools doesn't disqualify a book from KDP, doesn't affect royalties, and doesn't hurt ranking. What the policy actually requires is transparency about how AI was involved — and that "how" turns out to be the most important design decision in how you write your next book, not just a compliance checkbox at the end.
What Amazon Actually Requires
The rule is simple to state and harder to apply: you must inform KDP when your published book contains AI-generated text, images, translations, or other creative content, even if you subsequently edited it. Every book submission includes a section where you declare the role AI played. The declaration is internal to Amazon's systems — it is not displayed on your book's product page the way a fiction/non-fiction genre tag is — but it is mandatory, and it is checked.
Enforcement has tightened steadily. What started as a loosely worded content guideline in 2023 has become a specific, binary disclosure requirement with real consequences attached: undisclosed AI-generated content that should have been flagged violates KDP's terms of service, and Amazon has shown it will block listings or suspend accounts over it, not just quietly bury the book in search results.
The part that trips people up is that Amazon didn't just ask "did you use AI, yes or no." It split AI involvement into two categories, and the line between them is where all the actual judgment calls live.
AI-Generated vs. AI-Assisted: The Distinction That Decides Everything
This is the entire policy in one sentence: AI-assisted work does not need to be disclosed. AI-generated content does.
AI-assisted covers the AI use every writer has been doing for years without a second thought — brainstorming ideas, checking grammar, refining sentences you already wrote, cleaning up a human-created image. The defining feature is that a person originated the content and AI helped shape or polish it. Amazon does not require disclosure here, and for most professional non-fiction authors, this is where the majority of their AI use has always lived: an editing pass, a phrasing suggestion, a structural sanity check.
AI-generated covers content where an AI tool produced the substance — the text, the image, the translation — even if a human touched it up afterward. The test isn't "did I edit this at all," it's "who actually generated the material in the first place." Feed a topic and a one-line prompt into a chatbot, get back three chapters, lightly proofread them, and publish: that's AI-generated, full stop, regardless of how much proofreading happened at the end. The disclosure requirement follows the origin of the content, not the amount of human polish layered on top of it.
Where this gets genuinely difficult is the space in between, which is where most real books actually live:
- You outline a chapter yourself, ask an AI to draft a first pass from your outline and notes, then substantially rewrite it in your own voice. Generated or assisted?
- You write every sentence yourself but ask AI to restructure paragraphs for flow. Generated or assisted?
- You dictate ideas conversationally, AI turns them into prose, you edit for accuracy but keep most of the sentence structure. Generated or assisted?
Amazon's own guidance leans toward origin, not final-draft resemblance: if AI produced the first version of the words, that's generation, even when heavy editing follows. The safest read for authors is to disclose whenever AI produced a complete passage from a prompt, and treat "AI helped me write faster" and "AI wrote it and I revised it" as two different sentences describing two different processes — because Amazon is now asking you to tell them apart in writing.
How the Disclosure Actually Works
The mechanics are less dramatic than the stakes suggest. During the KDP submission flow, you'll hit a content declaration step asking you to categorize the AI involvement in your book — text, images, and translations are asked about separately, since a book can mix categories (human-written text with an AI-generated cover, for instance). You select the category that applies, submit, and the book goes through normal review.
A few things worth knowing before you get there:
- It's not public-facing. Readers don't see an "AI-generated" badge on your book's detail page. The disclosure is a compliance record between you and Amazon, not marketing copy or a warning label for shoppers.
- It doesn't gate approval by default. Declaring AI-generated content doesn't get your book rejected. Amazon's stated position is that AI content is allowed; the requirement is knowing about it, not preventing it.
- It's retroactive-unfriendly. If you've published books before this policy tightened and didn't disclose AI involvement that should have been flagged, that's a gap worth closing proactively rather than waiting for an automated content scan to flag it for you — and by 2026, automated detection has gotten good enough that "no one will notice" is a bad bet.
Why This Matters Beyond Compliance
The paperwork is the smaller issue. The larger one is what the AI-generated category signals about the book itself, and it has nothing to do with Amazon's terms of service.
Readers have gotten fast at spotting the pattern: the flattened rhythm, the hedge-everything phrasing, the examples that feel assembled rather than lived. That "AI voice" isn't a myth invented by skeptics — it's a real, recognizable output pattern from prompting a general-purpose model with a one-line instruction and publishing what comes back. As the volume of one-shot generated non-fiction on Amazon has grown, so has reader (and reviewer) skill at identifying it, and the market for that category of book is getting more crowded and more commoditized by the month. A glut of interchangeable AI-generated guides on the same topic doesn't help any single one of them sell.
That's the practical argument for staying on the AI-assisted side of the line even where you legally could disclose and publish AI-generated content: not because generation is against the rules, but because a book built from your actual expertise — your case studies, your framework, your specific point of view — reads differently from a book assembled by a model guessing at the average version of your topic. The disclosure category and the quality signal happen to point the same direction.
A Practical Checklist for Non-Fiction Authors
Before you submit, it's worth running your manuscript through a short gut-check:
- Did I originate the structure and argument, or did AI? If you built the outline and the chapter logic and AI helped draft or polish within that structure, you're closer to assisted. If you handed over a topic and got back a structure, you're closer to generated.
- Do the examples and case studies come from your actual experience, or were they invented by the model? Fabricated-but-plausible case studies are a strong generated signal, disclosure question aside — verify anything a model handed you before it goes in a published book.
- Would I recognize my own voice in this chapter if I read it cold? If not, an AI produced more of the substance than you did, whatever the eventual editing pass looked like.
- Am I disclosing consistently across text, images, and translations? A cover generated by an AI image tool and a translated edition run through machine translation are separate disclosure questions from the body text — don't assume one covers all three.
- If I published under the old, looser guidance, have I reviewed those books against the current rule? Tightening enforcement means older gaps are increasingly likely to surface.
None of this requires avoiding AI. It requires being honest with yourself about which side of the origin line your process actually falls on, before Amazon's declaration form asks you to be honest about it in writing.
Does This Extend Beyond Amazon?
KDP is the disclosure policy most authors run into first, because it's the largest self-publishing platform by a wide margin, but it isn't operating in isolation. IngramSpark and Draft2Digital have moved toward similar content declarations for wide-distribution titles, and the trend line across retailers points the same direction Amazon's already gone: transparency requirements, not prohibitions, with enforcement that has been getting stricter rather than looser as detection tools improve.
That matters if you distribute wide instead of exclusively through KDP. A book cleared through one platform's disclosure process isn't automatically cleared everywhere — each retailer runs its own declaration, and the categories don't always map one-to-one. The practical takeaway is the same regardless of where you publish: keep an honest internal record of how AI was involved at each stage of a manuscript — outlining, drafting, editing, cover art, translation — so that whichever platform asks, you have an accurate answer ready instead of reconstructing the process after the fact.
Common Questions Authors Ask
Does disclosing AI involvement hurt my book's sales or search ranking? Amazon has stated the disclosure does not affect royalties or ranking, and it isn't shown on your product page. It's a compliance record, not a marketing signal readers see.
What if I only used AI for grammar and light editing? That's squarely AI-assisted. Grammar checking, phrasing suggestions, and polishing text you originated don't require disclosure under Amazon's current guidance.
What about book covers made with an AI image generator? Cover art is disclosed separately from body text. If an AI tool produced the image, that's an AI-generated image, even if you directed the prompt and selected among variations — the same origin-based logic that applies to text applies to art.
Do I need to disclose AI use for a translated edition? Yes, machine translation of your manuscript into another language falls under the AI-generated translation category, separate from whatever declaration you made for the original-language text.
Can Amazon tell if I don't disclose? Detection tooling has improved substantially, and enforcement has tightened through 2025 and into 2026. Treat "it probably won't get flagged" as a bet against a system that's getting better at this every quarter, not a safe assumption.
Where This Points for How You Write the Book
The disclosure policy is, inadvertently, a pretty good argument for a particular way of working: keep the author driving the structure and the judgment calls, and let AI handle the mechanical load of turning direction into prose — rather than routing a topic through a generic chatbot and hoping the output needs only light editing.
That's the model WriteABookAI is built around. You generate the chapter outline and reshape it before any drafting starts. You supply your own source material — the frameworks, case studies, and terminology that make the book actually yours — so chapters draft against what you know rather than what a general model guesses. You direct tone, revise chapters, and rewrite sections throughout, rather than receiving a finished manuscript from a single prompt.
That workflow doesn't make the disclosure question disappear — if you use AI to help draft, some of what you disclose will still reasonably fall on the "assisted" side of Amazon's line, and you should be accurate about it either way. But it does mean the book you end up with is built the way readers can tell the difference: your argument, your evidence, your voice directing the process from outline to final chapter, with AI doing the part that used to cost professionals months of blank-page time. See how the structure-first, author-directed workflow works end to end.
The disclosure requirement isn't going away, and it's likely to get more specific, not less, as AI-generated volume on Amazon keeps climbing. The authors who won't spend any time worrying about which box to check are the ones whose process already keeps them clearly on the right side of the line — because they were driving the book the whole way through, not assembling one from a prompt.