Do You Need To Disclose AI Use When Publishing A Book?

If you used any AI tool while writing your book — from an outline generator to a full drafting assistant like an AI ebook generator — you've probably wondered whether you're now required to tell someone. The honest answer isn't a single yes or no. It depends on which "someone" you mean: a retailer whose terms of service you agreed to, a government office you might register your copyright with, or the reader who's about to spend money on your book. Those are three different questions with three different answers, and conflating them is where most of the confusion — and most of the bad advice circulating in author forums — comes from.

This piece separates the three, gives you Amazon KDP's actual current wording rather than a paraphrase of a paraphrase, and lays out what to do with that information once you have it.

Three different questions hiding inside one

"Do I need to disclose AI use?" usually means one of these:

• Is there a law that requires it? In most jurisdictions, no general law requires a self-published author to tell readers they used AI to write a book. There's one narrow, specific exception worth knowing about, covered below, involving copyright registration — not book sales.
• Does my publishing platform require it? Increasingly, yes — and this is the one with real teeth, because violating a platform's terms of service can get a book removed or an account suspended, which is a much faster consequence than any hypothetical lawsuit.
• Should I tell readers, even if nobody's making me? This is an ethical and trust question, not a legal one, and it's the one most authors actually mean when they ask.

Treat those as separate checklists, not one blurry worry, and the whole topic gets a lot less stressful.

What Amazon KDP actually requires today

Amazon's Kindle Direct Publishing content guidelines contain a specific, current requirement on this, and it's worth quoting rather than summarizing secondhand: KDP states that it requires authors "to inform us of AI-generated content (text, images, or translations) when you publish a new book or make edits to and republish an existing book" through KDP.

That's a platform disclosure — made to Amazon, during the publishing workflow — not a requirement to put a disclaimer in your book's description or front matter. KDP's own guidelines don't spell out whether, or how, that internal answer is surfaced to shoppers on your book's product page; what's explicit is only that Amazon has to be told. If you want readers to know too, that's a separate, voluntary choice covered further down.

The more useful part of KDP's guidance is how it draws the line between two categories that sound similar but aren't treated the same way at all.

AI-generated vs. AI-assisted: the line KDP actually draws

This is the distinction the whole policy hinges on, and it's worth getting exactly right instead of approximating it.

AI-generated content, per KDP's guidelines, is "text, images, or translations created by an AI-based tool." The key sentence: if an AI tool produced the actual content, "it is considered 'AI-generated,' even if you applied substantial edits afterwards." That last clause matters more than it looks — it means heavily rewriting an AI draft doesn't reclassify it as your own unassisted writing for disclosure purposes. If AI wrote the first version, KDP treats it as AI-generated content, full stop, regardless of how many editing passes came after.

AI-assisted content is different, and doesn't require disclosure at all. KDP's guidelines describe this as using AI tools to "edit, refine, error-check, or otherwise improve" content you created yourself, and separately, using AI "to brainstorm and generate ideas, but ultimately" creating "the text or images yourself." Grammar checking, structural feedback, brainstorming a chapter list, running your own draft through an AI editor for line notes — none of that is AI-generated content under KDP's definition, because you wrote the words.

So the test isn't "did AI touch this book anywhere in the process." It's narrower: who wrote the sentence that ended up on the page — you, or the model? If a tool drafted a paragraph and you then rewrote it heavily, KDP still calls that AI-generated, because the tool produced it first. If you drafted the paragraph and a tool tightened your sentences, that's assisted, and it's yours.

Practically, this means a huge share of ordinary AI-assisted writing work — outlining with an AI ebook generator, using AI to catch inconsistencies, asking a model to suggest better transitions between two paragraphs you wrote — sits on the "no disclosure needed" side of KDP's own line. What crosses into the "must disclose" category is specifically the AI-drafted content that made it into the final book, whether that's a chapter an AI tool wrote and you edited, AI-generated cover art, or an AI translation of your text into another language.

Being responsible for what AI produced, disclosed or not

KDP's guidelines pair the disclosure requirement with a separate, broader responsibility clause: authors are responsible for verifying that all AI-generated and AI-assisted content complies with KDP's content guidelines, including intellectual property rights, and are told to review generated content thoroughly because it "may include information that wasn't in your original input." In other words, disclosure and accuracy are two different obligations that both apply — telling Amazon a chapter was AI-generated doesn't excuse you from fact-checking it, and fact-checking it doesn't excuse you from disclosing it. Both boxes need checking, not one or the other.

That's also a useful reminder of why disclosure exists on KDP's side at all: it's less about warning readers and more about Amazon needing to know, at scale, which books in its catalog came from a process it has separate quality and rights-compliance concerns about — the same reason a retailer's terms of service exist for any category prone to abuse.

What happens if you skip it

KDP's guidelines are direct about this part: content that violates the guidelines, disclosure included, is subject to removal, and repeated or serious violations can affect your publishing account. That's not a hypothetical — it's the stated consequence for content guideline violations generally, and AI-generated content that goes undisclosed is a content guideline violation like any other. Given how fast AI-detection tooling on the retailer side has improved, and how much scrutiny AI-flooded categories have drawn from platforms trying to protect catalog quality, treating the disclosure question as "nobody will notice" is a worse bet in 2026 than it might have looked a couple of years ago. Answer the question KDP asks honestly; it costs nothing and the downside of guessing wrong is a book pulled from sale.

The one place disclosure is an actual legal requirement

Here's the exception that's easy to miss because it has nothing to do with retailers: the U.S. Copyright Office. If you register a book for copyright and it contains AI-generated material, the Copyright Office's registration guidance requires you to disclose that AI-generated content in the application and to describe the human author's contribution, so the AI-generated portions can be excluded from the claim. As the Authors Guild's summary of the Copyright Office's guidance puts it, authors should "disclaim" AI-generated text in a registration "just as they would" for any other material they didn't personally create.

The practical stakes: undisclosed AI content in a copyright registration doesn't just risk a rejected application. Legal commentary on the guidance — including analysis from IP firms covering the same Copyright Office report — notes that a registration that omits a required AI disclaimer is vulnerable to cancellation later, and can hand an infringer or opposing party grounds to challenge the registration's validity in court. That's a real, binding legal obligation with real consequences, distinct from anything KDP or any retailer asks for — it's about what you can claim exclusive rights to, not about what a bookstore lists on a page. Purely AI-generated brainstorming, outlines, and research don't affect copyrightability; the requirement kicks in specifically for AI-generated expression that ends up as text or images in the finished, registered work.

Most self-published authors never think about this because most never file a standalone copyright registration — U.S. copyright exists automatically at the moment of creation, registration is a separate optional step usually pursued for stronger legal remedies. But if you do register, or if a publishing contract requires you to, this is the one context where "disclose the AI use" stops being a platform courtesy and becomes an actual legal filing requirement with real teeth.

How other platforms and retailers currently handle it

KDP isn't the only publishing venue setting rules here, and the rules aren't uniform — which matters if you plan to distribute the same book to more than one storefront.

| Platform | Current stance on AI-generated content | |---|---| | Amazon KDP | Requires disclosure to Amazon of AI-generated text, images, or translations at publish/republish time; AI-assisted content (editing, brainstorming) is exempt from disclosure | | IngramSpark | Its Catalog Integrity Guidelines list as prohibited "content created using automated means, including but not limited to content generated using artificial intelligence or mass-produced processes" | | Barnes & Noble Press / Draft2Digital | Both platforms require all submitted content to meet their content policies regardless of AI involvement; both have tightened account and title limits in response to a wave of low-effort AI-generated submissions flooding their catalogs |

The direction of travel across all of them is the same even where the specific rule differs: platforms are getting stricter, not looser, about AI-generated content as volume increases, and "check the current policy on the specific platform you're publishing to" is better advice than assuming one retailer's rule applies everywhere you distribute. A book that's fine under KDP's disclose-and-publish model could run into a much harder line elsewhere. If you're publishing wide, read each platform's current policy yourself before assuming the rules match.

Why "no legal requirement" isn't the same as "no reason to tell readers"

Strip away the platform and copyright-office questions and you're left with the one that actually decides how most authors handle this: no law and, outside a KDP-style storefront requirement, no contract forces you to tell a reader you used AI. So why do it anyway?

The Alliance of Independent Authors, a nonprofit membership organization for self-published authors, has published ethical guidelines for authors using AI on exactly this question, built around principles it summarizes for authors as clarity, curiosity, creativity, and accountability. Its framing is direct: "even where the law does not require transparency, ethics often do." ALLi doesn't ask authors to over-disclose — its guidance explicitly notes that authors shouldn't feel obligated to label every minor AI-assisted task on a copyright page — but it does frame a clear, honest AI policy as something closer to a professional norm than an optional extra, describing it as a trust signal that shows readers respect rather than a confession that undermines the book.

That framing is worth sitting with, because it reorients the question away from "am I forced to say this" toward "what actually protects the relationship I'm building with readers." A reader who buys a memoir expecting a person's real, lived voice and later learns AI drafted large sections of it isn't upset because a rule was broken — often none was. They're upset because what they bought didn't match what they thought they were buying. Disclosure, done well, prevents that mismatch before it happens instead of managing the fallout after a reader figures it out on their own, which — given how visibly AI-generated prose has started showing up in reviews and reader discussion — is a more common discovery path than authors relying on silence tend to assume.

Why claiming "100% human-written" is the riskier move

There's a version of this question authors don't ask often enough: not "should I disclose," but "what if I actively deny it." Some authors, worried that any AI mention will tank sales or credibility, go a step further than staying quiet — they explicitly market a book as entirely human-written when AI played a real role in producing the text. That's a meaningfully bigger risk than honest disclosure, for a few compounding reasons.

First, it converts a disclosure question into a false-statement question. Not telling a reader you used AI, where no rule requires you to, is a values call. Telling them you didn't, when you did, is a claim you're making that isn't true — and on a platform like KDP, where the disclosure question is a required part of the publishing workflow, answering it falsely isn't a gray area at all; it's a direct violation of the terms you agreed to, with the same removal and account consequences as any other content guideline breach, except now it's compounded by having affirmatively lied rather than simply stayed silent.

Second, AI-detection tools aimed at published text keep improving, and they're being used by readers, reviewers, and platforms alike. A false "100% human" claim is a claim that can be checked, and increasingly is — sometimes by a single skeptical reviewer with a detector tool and a grudge, which is a much lower bar to clear than "nobody will ever notice." A book quietly published with no disclosure statement either way faces relatively low scrutiny risk. A book that loudly claims to be entirely human-written and then gets caught being otherwise faces a specific, avoidable reputational hit — the story becomes "this author lied," which is a worse outcome for trust than "this author used a tool," even among readers who care about AI use.

Third, and most simply: honest disclosure has a ceiling on how bad it can go — some readers won't buy the book, which was probably true before they knew anyway. A false claim has no ceiling, because the failure mode isn't "fewer sales," it's "credibility problem that follows the author to their next book, and the one after that." If you're weighing the risk of saying too much against the risk of claiming too much, the second one is structurally worse in every scenario except the one where you never get caught — and that's a bet, not a strategy.

How to disclose without over-explaining

None of this means every book needs a lengthy AI methodology essay in the front matter. Over-disclosure has its own cost — it can read as defensive, or worse, as marketing copy dressed up as transparency. A few practical patterns that stay honest without turning into a disclaimer nobody reads:

• Match the disclosure to the actual role AI played. A book where AI drafted full chapters warrants a different note than a book where AI only helped with brainstorming and line edits — the second doesn't need a disclosure at all under most platforms' own definitions, so don't manufacture one out of caution.
• A single sentence usually does the job. Something as short as "This book was written with the assistance of AI writing tools, then researched, edited, and finalized by the author" tells a reader everything relevant without turning into an essay. Save the longer version, if you want one, for your website or an author's note rather than the book's back cover copy.
• Be specific about what AI didn't do, if that's the part you're proud of. If the research is yours, the stories are yours, and the final judgment calls are yours, saying so is more useful to a reader than a vague "AI helped" — it tells them what to trust and what to weigh, rather than just that a tool was somewhere in the process.
• Put it where a genuinely curious reader would look, not where it's hardest to find and not where it's impossible to miss. A line in the front matter or an "about this book" section on your author site both work; burying it in terms nobody reads defeats the purpose, and plastering it across the cover is rarely proportionate to what actually happened.
• Keep your platform disclosure and your reader-facing note consistent. If you told KDP a chapter was AI-generated, don't write reader-facing copy that implies otherwise — the inconsistency, if anyone compares the two, is worse than either statement alone would have been.

The goal isn't performing transparency. It's giving a reader accurate information about what they're buying, in proportion to how much AI actually shaped the finished book — which for most authors using a tool like an AI ebook generator as a drafting and outlining aid, then doing real editorial and fact-checking work on top of it, is a shorter and less dramatic disclosure than the anxiety around the question usually suggests.

A simple way to decide, book by book

Run through this in order and the right move for a specific book usually becomes obvious:

1. Did an AI tool produce the actual text, images, or translations that ended up in the finished book — not just brainstorming or editing, the content itself? If no, you're in AI-assisted territory on most platforms' own terms, and formal disclosure likely isn't required at all.
2. If yes, does the platform you're publishing to require disclosure? Check its current policy directly rather than assuming — KDP does, with the AI-generated/AI-assisted distinction above; other platforms vary, and some prohibit AI-generated content outright rather than just requiring disclosure.
3. Are you planning to register the book's copyright? If so, the AI-generated portions need to be disclosed and disclaimed in that specific application, separate from anything a retailer asks.
4. Even where nothing above requires it, would a reader feel misled if they found out later how the book was actually written? If the honest answer is yes, that's the strongest signal to disclose voluntarily — not because a rule says to, but because the alternative is a reader discovering it on their own and drawing worse conclusions than an upfront sentence would have produced.

None of these four questions has a scary answer for an author who used AI as a genuine drafting and research aid and then did the real editorial work of fact-checking, revising, and making the book actually theirs. The scary version of this topic mostly belongs to authors trying to hide that process entirely or claim credit for none of the tool's involvement — and as covered above, that's the position with the worst downside, not honest disclosure. Get the process right, tell the platform what it asks for, and a short, honest note to readers costs you far less than either silence discovered later or a false claim that doesn't hold up.