Amazon KDP's content guidelines say authors must disclose AI-generated content — text, images, or translations — "even if you applied substantial edits afterwards." That sentence is doing a lot of work, and it's the reason KDP's own community forums are full of authors asking versions of the same question: does what I did count?
The rule itself doesn't define "substantial," doesn't say what counts as "AI-assisted" versus "AI-generated," and doesn't walk through common real-world workflows. Here's how the boundary actually falls, based on Amazon's published guidance and the disclosure categories it asks authors to choose between at upload.
| Category | What it means | Disclosure required? |
|---|---|---|
| AI-generated | An AI tool produced the content (text, image, or translation) with minimal or no human authorship, even if you edited it afterward | Yes |
| AI-assisted | You wrote the content yourself; AI tools helped with brainstorming, editing, grammar-checking, or research along the way | No — but see the note below |
| No AI | No AI tool was used anywhere in the content's creation | No |
"I wrote a rough draft, then had AI rewrite it heavily for style." This is the single most common gray area, and it's exactly where the "substantial edits" language bites: if the AI tool is the one producing the final prose — even starting from your own draft — that output counts as AI-generated text, not AI-assisted. The test isn't how much of the ideas were yours; it's whether the final words on the page were generated by the tool.
"I used AI to fix grammar and tighten sentences, nothing more." This is the textbook AI-assisted case and doesn't require disclosure under KDP's own categories — the same way spell-check or a human copyeditor wouldn't. The line is whether the tool is generating new content or correcting content you already wrote.
"My cover artist used an AI tool as part of their process, but I don't know the details." If AI-generated imagery ends up in your cover or interior art, it needs disclosure regardless of whose workflow produced it. If you commissioned cover art, it's worth asking the artist directly whether AI tools were involved in the final image, not just early concepting.
"I translated my own English book into Spanish using an AI translation tool." AI-generated translations are their own disclosure category, separate from text and images — this applies even if the original-language manuscript required no disclosure at all.
"I'm not sure which category my book falls into." When a workflow mixes AI-generated passages with human-written ones — common in long manuscripts — the safer read of "even if you applied substantial edits afterwards" is to disclose. Under-disclosing is the version of this mistake that carries account risk; over-disclosing does not.
KDP has moved from a self-reported honor system at upload time toward retroactively reviewing previously published titles. A book that went live months ago without disclosure isn't necessarily in the clear — if it's flagged later, the consequence isn't just an edit request, it can include account-level enforcement. That retroactive risk is the reason "what counts" is worth getting right before publishing, not something to reason about after the fact.
None of this is a reason to avoid using AI tools in your workflow — it's a reason to know, in plain terms, which of your specific choices need a checkbox and which don't.
Get your book's exact disclosure requirement — free →