KDP introduced AI content disclosure in late 2023. Three years later, the policy is more enforced but not more clear. Here is what actually changed each year, what is still undefined, and what you need to do about it.
Timeline Overview
| Period | What changed |
|---|---|
| September 2023 | KDP introduces AI content disclosure checkbox in upload flow |
| Late 2023 | Initial confusion over what counts as "AI-generated" vs "AI-assisted"; no clear KDP definition |
| 2024 H1 | KDP updates Help Center with clearer AI-generated vs. AI-assisted guidance |
| 2024 H2 | Disclosure requirement confirmed to cover images and translations, not just text |
| 2025 | Enforcement activity increases; suspensions reported for undisclosed AI content; explicit policy warning added for circumvention |
| 2026 (current) | No major new policy changes; 2024 framework in effect; enforcement continues |
2024: The Defining Year — AI-Generated vs. AI-Assisted
The most important clarification in KDP's AI policy history came in 2024: the line between AI-generated (disclosure required) and AI-assisted (no disclosure required).
What KDP Defined as AI-Generated (Disclosure Required)
| Content type | AI-generated example |
|---|---|
| Text | AI wrote the chapter, scene, or article — human edited afterward |
| Images | AI image generator created the book cover or interior illustrations |
| Translation | AI translated the book from another language |
| Narration / audio | AI voice synthesized the audiobook narration |
What KDP Defined as AI-Assisted (No Disclosure Required)
| Content type | AI-assisted example |
|---|---|
| Text | Human wrote the draft; used AI for grammar/spell-check |
| Text | Used AI to research topics or outline; human wrote the content |
| Text | Used AI to suggest edits; human accepted or rejected changes |
| Images | Used AI to remove background from a photograph |
| Translation | Human translated; used AI to check phrasing |
The key test KDP applied: Who authored the content? If a human authored it and an AI tool helped polish it, that is AI-assisted. If an AI tool authored it and a human edited the AI's output, that is AI-generated.
This distinction was not formalized in a single published document — it appeared in KDP Help Center articles and in KDP's responses to author questions via support.
2025: Enforcement, Not New Rules
2025 did not bring new policy categories — it brought enforcement of the 2024 rules.
Suspensions for undisclosed AI content: Author communities (SFWA, ALLi, Kboards) documented KDP account suspensions and title removals for AI-generated content that was not disclosed. KDP's detection methods appeared to include image generation metadata, content pattern matching, and information from disputes.
Anti-circumvention language added: KDP made explicit that attempting to disguise AI-generated content as human-authored violates the ToS. The consequence is account termination, not just content removal.
Still no threshold: KDP did not publish a minimum percentage. The practical implication: if any material portion was AI-generated, disclose. There is no "it was only 20% AI" defense.
Kindle Unlimited quality reviews: Reports of increased quality screening for AI-generated content in KU. KDP did not publish a policy — some AI-heavy content was removed for quality violations, not AI disclosure violations specifically. The distinction matters: a quality violation removes the book; a ToS violation can remove the account.
2026: Current State
As of May 2026, the KDP AI policy is:
| Requirement | Current status |
|---|---|
| Disclosure checkbox at upload | Required for AI-generated text, images, translations |
| AI-generated definition | AI authored the content (editing by human does not change this) |
| AI-assisted definition | Human authored; AI helped edit, check, suggest |
| Enforcement | Active; undisclosed AI content can result in removal or suspension |
| Kindle Unlimited eligibility | Not explicitly restricted based on AI disclosure — but quality standards apply |
| Royalty rates | Unchanged based on AI disclosure |
What KDP still does not address in 2026:
- What percentage of AI content in a book triggers disclosure (all-or-nothing is implied)
- Whether KDP will ever restrict AI-generated content from certain categories (children's books, educational content)
- Whether disclosed AI content faces different merchandising, ranking, or recommendation treatment
- How KDP handles AI-generated content that was uploaded before the disclosure requirement existed
What Has Not Changed Since 2023
Despite three years of AI content policy evolution, several things have stayed the same:
- Disclosure does not equal approval — checking the AI box does not guarantee KDP will keep your book live if it violates other content or quality policies
- Quality standards apply regardless — AI-generated content that is low-quality, repetitive, or spammy can be removed for quality violations even if properly disclosed
- No AI-generated CSAM — prohibited absolutely; disclosure does not create any exception
- Human accountability — the author of record remains responsible for the content regardless of how it was produced
What to Do Now
If you have existing titles on KDP:
- Audit your catalog — for each title, determine whether the text, images, or translations were AI-generated (not just AI-assisted)
- Retroactively disclose if needed — KDP allows updating content metadata; if you have undisclosed AI-generated content, disclose it proactively rather than waiting for a review
- Document your production process — maintain notes on what tools were used and whether the human or AI authored each material element; this protects you if KDP questions a title
- Check Help Center for updates — KDP policy can change via Help Center updates without press announcements; check kdp.amazon.com/help before uploading new AI-assisted work
Sources: Amazon KDP Help Center, KDP content guidelines (multiple versions), author community reports (SFWA, ALLi, Kboards). This article reflects KDP policy as of May 2026; KDP can change policies via Help Center without formal announcement.
