What Should Self-Publishers Disclose to KDP When Using AI in 2026?

KDP AI Disclosure in 2026: The Direct Answer

Amazon KDP does not require authors to disclose ordinary spelling, grammar, brainstorming, fact-checking, formatting, or other AI assistance. Its published guidance requires disclosure of AI-generated text when publishing a book through KDP, particularly when the text is published in the author’s name. This includes books created wholly or partly with generative AI when that generated material forms part of the work being distributed, such as a passage, poem, story, or substantial section. Merely using AI to organize notes, suggest a title, tighten wording, or generate a cover may not trigger the same disclosure requirement. As of 27 September 2026, authors should treat the boundary as material authorship: if AI produced words that readers will reasonably think the author wrote, disclose it. Transparency does not automatically make a title ineligible for KDP; concealing generated text is the larger compliance problem.

Also worth reading: What is an AI cover art compliance checklist for self-publishers in 2026? · What are the KDP AI disclosure requirements for self-publishers on Amazon? · How Should Publishers Build AI Editorial Governance in 2026?

The disclosure does not mean an author must say that every sentence was AI-generated. A concise statement such as “Parts of this book were created with generative AI and reviewed and edited by the author” can communicate the material use without turning the front matter into a technical report. Authors should also avoid implying that a human wrote every word if the opposite is true. Disclosure is separate from copyright ownership, quality control, and platform eligibility: KDP can accept AI-assisted work, but the author remains responsible for accuracy, rights, metadata, and the promise made on the product page. The safest approach is to inspect the book, identify meaningful AI-generated passages, and use plain language that does not overstate or minimize the role of the tool.

What KDP’s AI Content Rules Actually Cover

KDP’s content guidelines distinguish AI-generated text from AI-assisted work. The platform’s stated position is that authors must disclose AI-generated content, not necessarily content for which AI only provided support. A writer who uses a chatbot to propose 20 chapter ideas remains the author if the writer plans, drafts, and edits the manuscript. A writer who pastes substantial generated prose into the file and submits it without identifying that use may be making a materially different representation. The rule is designed around content published under an author’s identity, not around a permanent ban on using software. That distinction explains why a grammar checker does not require disclosure while generated narrative passages generally do.

Disclosure also does not erase an author’s obligations under KDP’s content guidelines. The publisher must have rights to all text, images, fonts, trademarks, and other material in the book. AI output can reproduce recognizable writing, imitate a named creator, or create accidental similarities, so generated material should not be accepted merely because the model produced it quickly. Authors should compare passages against their sources, replace accidental close matches, and avoid prompts designed to reproduce protected material. They must also verify names, dates, quotations, legal claims, and biographical statements. Amazon may remove content, restrict accounts, or refuse payment where its policies or applicable law are violated, regardless of whether an AI disclosure appears.

There is no generally published KDP rule requiring authors to upload a separate “AI certificate,” disclose the model name, or provide the exact percentage of generated text. Nor does Amazon generally require public labeling of a book merely because AI helped with spelling, grammar, keywords, or cover ideation. The relevant issue is truthful authorship and compliance. Because policies can change, authors should recheck the current Content Guidelines and Publishing Information before uploading a title, and should document which tools were used and which passages received material AI assistance.

AI Disclosure Checklist for KDP Authors

The practical KDP AI disclosure checklist is based on a simple sequence: identify, examine, document, disclose, and retain evidence. First, identify every generative tool used for manuscript development, including web-based chatbots, image generators, and writing applications with built-in generation features. The author should record the service used, the date of use, the purpose of the request, and the parts of the book affected. This record should include cover and interior images if they were generated. A spreadsheet is usually enough, and the record may be more detailed than the public statement. Its purpose is to prevent an author from forgetting a generated section when preparing the KDP manuscript or answering an account inquiry.

Second, the author should classify each use as assistance or material generation. Assistance includes outline suggestions, plot brainstorming, grammar correction, consistency checks, and keyword research. Material generation includes drafted scenes, essays, explanations, poems, summaries, character dialogue, or rewritten passages that substantially originated with a model. The labels are not perfectly objective. If an author requests a paragraph, accepts it unchanged, and then inserts it into the final manuscript, that is a strong case for disclosure. If the writer uses a model as a critic, rejects the output, and independently creates the final prose, disclosure may not be required under the stated policy, although voluntary transparency can still be sensible.

Third, the author should review the complete file rather than assuming that only obvious AI wording counts. Generated text can sit in introductions, appendices, sample material, author notes, and captions. It can also appear in a children’s activity page or a nonfiction summary while the narrative chapters were written conventionally. Search the manuscript for boilerplate phrases, generic transitions, unsupported claims, repeated rhythm, and references to “as an AI,” then correct any language that accidentally exposes the drafting process. Finally, the author should place a clear disclosure where readers and KDP reviewers will find it, while also considering whether the KDP product-details section needs an explanatory note under the platform’s current instructions.

How to Write a Clear and Proportionate Disclosure

A useful disclosure should be short, factual, and understandable to a reader who has never used generative AI. “Parts of this book were created with generative AI. The author reviewed, edited, and takes responsibility for the final content” is usually clearer than an acronym-filled paragraph naming several tools. The wording should identify the affected material when readers might otherwise assume fully human composition. A memoir containing an AI-generated introduction could say, “The introduction includes generative-AI-assisted text that was reviewed and edited by the author.” A fully generated novel may require a more direct statement, such as, “This novel was created with generative AI, and the author edited and arranged the final manuscript.”

The disclosure should not claim that AI “wrote nothing” if material passages were generated, but it also should not exaggerate a minor support role into an admission that every word came from a model. Authors should avoid vague wording such as “AI was involved,” because that can leave the most important fact unresolved. If the book is fiction, authors may use “generative AI” rather than “ChatGPT,” particularly if several services were involved. If images or synthetic voices were used, the disclosure and copyright records should address those elements separately because text disclosure does not automatically settle rights in media.

Placement depends partly on the format. Most authors should consider front matter, the copyright or edition page, and any relevant KDP metadata field. A disclosure buried in an author biography may be too late or too obscure, while a dramatic cover badge can unnecessarily dominate a commercial design. Keep the explanation proportionate: one or two sentences are often enough. Authors should also ensure that the disclosure remains accurate after later edits, because removing an AI-generated passage may change whether disclosure is still necessary. The best wording is the narrowest statement that truthfully describes the final edition.

KDP, Traditional Publishing, and Other Distribution Options

No mainstream platform has adopted exactly the same AI policy. KDP’s requirement to disclose AI-generated text is important, but authors comparing options should examine four separate issues: disclosure, rights, media rules, and account enforcement. Traditional publishers often have submission policies and contract terms that address written, visual, audio, and other synthetic material, but a contract may permit more flexibility while still requiring disclosure to an editor. Some services may permit generated content, some may restrict it, and some may have no explicit public rule. Authors should not treat another platform’s silence as permission to upload undisclosed material on KDP.

The comparison also applies beyond ebooks. Audiobook narration, cover art, translated editions, and translated text may each have separate platform requirements. KDP’s written-content policy should not be treated as a universal permission for generated covers, cloned voices, or automatic translations. A tool may produce material that is original-looking yet still infringe copyrights, publicity rights, privacy rights, or contractual restrictions. Authors should select tools with suitable commercial terms, avoid impersonation, and retain a record of licenses and source materials. The distribution channel does not transfer responsibility to the service; the account owner still makes the representations at upload.

Costs, Royalties, and the Business Effect of Disclosure

Disclosure itself normally has no direct fee on KDP. The real costs are time, review, editing, rights research, and any tools the author purchased. Generative AI subscriptions vary widely, but authors should separate subscription expense from one-time publication costs. A professional edit can be priced by word count or project, and a copyright lawyer or rights investigator may charge substantially more than a grammar service. The largest financial risk is publishing inaccurate or infringing material that is later removed, delayed, or charged back. A transparent workflow may cost a few hours upfront and reduce the chance of an account-wide dispute.

KDP royalties make this distinction economically important. For eligible ebooks priced from $2.99 to $9.99, authors commonly receive a 70% royalty share after delivery costs, while the standard share outside that range is 60%. With KDP Select, royalties are generally paid at 70% for the first 90 days of a member’s paid-registration period and 30% thereafter. Kindle Unlimited page reads are funded through a fund and are not converted automatically into a fixed royalty per read, so an author should model that uncertainty rather than promise a precise income from AI-assisted production. Disclosure is not a reason to publish cheaply, and speed is not a substitute for reader value.

Print costs follow another model. A paperback’s list price and print cost depend on trim size, paper type, page count, color choices, and marketplace region. KDP’s print royalty generally equals 60% of list price minus printing cost, subject to its current pricing rules and regional conditions. These economics reward editing and proofreading: a generated manuscript with factual errors can lose money if readers return it or the book needs a costly correction. Authors should compare the labor they would otherwise perform with professional services rather than treating AI as a guaranteed margin improvement. A disclosure that attracts criticism is less damaging than a quality failure traceable to unverified generation.

Common Mistakes and Account Risks

The first common mistake is treating disclosure as a substitute for permission. A sentence saying “this book used AI” does not authorize copied text, trademarked logos, private facts, or a synthetic voice. The second mistake is assuming that a human editor automatically converts generated text into conventional human writing. An editor may correct grammar without rethinking a factual claim or rewriting a substantial passage, so the author must still inspect the final file. The third mistake is using disclosure as a shield against poor quality. KDP customers still expect readable prose, accurate promises, and a coherent product, and obvious generic language can damage both sales and reviews.

Authors also make the opposite error by disclosing every minor use in a way that misrepresents the book. Repeated warnings about a spelling checker can distract from the actual issue and mislead readers into believing that the manuscript was fully machine-written. Authors should apply a material-use test and describe the role accurately. A separate record can contain fine-grained details, while the public statement can be concise. Another error is changing the manuscript after upload without checking whether the disclosure and rights statement still match. A new edition may add generated material, so the workflow should be repeated whenever substantive text changes.

Finally, authors should not rely on screenshots, expired policies, or advice from an unrelated marketplace. The relevant source is the current KDP Content Guidelines, Publishing Information, and applicable Amazon terms, checked on 27 September 2026 and again before a later publication. Keep copies of the manuscript, source notes, licenses, prompts where appropriate, and the final disclosure. If Amazon contacts the account holder, a calm factual response is better than a contradictory explanation. Proportionate disclosure, documented permissions, and human verification are the strongest defense against both policy and reader complaints.

When Authors Should Act Before Publishing or Updating a Book

Authors should perform this review before uploading any new title, not after a complaint or a takedown. That includes fiction and nonfiction books, poetry collections, activity books, anthologies, educational supplements, and books assembled from licensed or generated source material. If several collaborators are involved, name the person responsible for each contribution and confirm that everyone understands the disclosure. If the author cannot identify which passages were generated, that uncertainty is itself a reason to pause. An author can replace the questionable material, rewrite it independently, or disclose the use; publishing while unsure creates avoidable risk.

The review is also warranted when a tool changes, the manuscript receives a major revision, or a new language edition is produced. Translation deserves particular attention: machine-translated prose may be generated text even when the source manuscript was human-written. Authors should compare a sample, check terminology, and confirm whether KDP treats the translated edition as a new work requiring updated information. Synthetic narration or cover art should be reviewed under the relevant media rules rather than folded into a generic text disclosure. Record the service terms in force when the material was produced, because later terms may not prove what the author was permitted to do at that time.

Timing matters for readers as well as compliance. Disclosing only after launch can look like an attempt to avoid scrutiny, while a vague statement at upload can be missed by buyers. The best moment is before the product goes live, when the manuscript and metadata can still be corrected. For an already published book, assess whether the current policy was violated or whether the disclosure is simply unclear. If the final text is unchanged and the existing description accurately identifies material AI use, the correction may be metadata and front matter; if generated text was concealed or rights are uncertain, pause distribution and seek help. Authors should not delete evidence or quietly replace material while an inquiry is open.

The final practical rule is straightforward: disclose material AI-generated text, disclose synthetic media when the relevant platform requires it, and verify that every part of the book is lawful and accurate. As of 27 September 2026, this approach aligns with KDP’s published distinction between AI-generated and AI-assisted work while avoiding the myth that all technology use is forbidden. It also gives buyers a truthful account of the book without claiming that AI guarantees quality. An author who documents use, reviews the output, and makes accurate statements is not merely protecting an account; the author is setting expectations for a product that deserves to sell on its merits rather than on ambiguity.