AI-Assisted vs AI-Generated: Where Amazon, Publishers, and Readers Draw the Line

Every author using any AI tool eventually asks the same nervous question: does this count as AI writing, and do I have to tell anyone? The line between AI-assisted and AI-generated work is not a vibe, it is a specific distinction that Amazon and publishers define, and getting it wrong can cost you your account.

Here is the distinction in one breath. Your book is AI-generated when an AI tool produces the actual text, images, or translation, even if you edit it afterward. Your book is AI-assisted when you write it yourself and use AI to brainstorm, outline, research, analyze, or suggest edits. Amazon requires you to disclose AI-generated material and does not require you to disclose the assisted kind. Who created the words is the whole test.

I have self-published eleven thrillers, and I treat every one like a product with a compliance checklist, because a blocked book earns zero dollars. So I read the actual policies instead of the panic threads. The good news is that the rule is more stable and more sensible than the fear around it suggests.

Key takeaways

• The test is authorship: if AI produced the prose, it is generated. If you produced the prose and AI helped around it, it is assisted.
• Amazon KDP requires disclosure of AI-generated text, images, and translations, and does not require disclosure of assisted work.
• The disclosure is internal to Amazon. It is not printed on your book's product page and Amazon says it does not affect royalties or ranking.
• Brainstorming, outlining, research, feedback, and edit suggestions all fall on the assisted side as long as you write the words.
• Enforcement tightened through 2025 and 2026. Non-disclosure can get a book removed, and repeat violations can end an account.

What Amazon actually requires

When you publish a new title on KDP, the AI question appears on the Content Details page, the second step of the workflow. It asks whether the content is AI-generated, with a checkbox for "Yes, some of this content is AI-generated." Tick it and Amazon asks a follow-up: which content type, text, images, or translation. That is the entire mechanism, and per Amazon's KDP guidance summarized here, the disclosure is used internally and is not displayed to customers on your listing.

The part that calms most of my author friends down is that the assisted category is broad. Amazon treats using AI to brainstorm ideas, build outlines, research a topic, or refine and edit prose you wrote as assistance, not generation. If you wrote the sentences, you are not ticking the box. If a tool wrote the sentences and you polished them, you are. The distinction survives your editing pass, which is the detail people miss: lightly rewriting AI output does not convert it to human-written work in Amazon's eyes.

Why the distinction is not just Amazon's rule

Traditional publishers and major magazines have landed in roughly the same place, and so have readers, which is why this matters beyond one retailer's checkbox. Most publishing contracts now include a clause requiring the author to confirm the work is human-authored, with AI permitted as a tool rather than a ghostwriter. Literary magazines that ban "AI writing" almost always mean generated prose, and are quietly fine with a writer who used a tool to catch continuity errors.

Readers draw the line in the same spot, only more emotionally. The backlash you see online is aimed at books that read like a machine wrote them, not at writers who used software to check their pacing. Nobody has ever left a one-star review because an author ran a readability report. They leave one-star reviews when the prose has that flat, average, nobody-actually-meant-this quality that generated text still carries. Staying on the assisted side of the line is not just compliance, it is how you keep the thing readers are actually paying for, which is you.

A practical map of common activities

The reason this question feels murky is that "using AI" covers a dozen very different activities. Here is how I sort them before I publish, so I know exactly what, if anything, I disclose.

The pattern is clean once you see it. Analysis, feedback, and planning are assistance. Production of the actual creative content is generation. When I want a tool that lives entirely on the assisted side by design, I reach for analysis-first software like storywith.ai, which reads the manuscript I wrote and tells me where it is weak without ever drafting a word for me. That keeps my compliance answer simple, and it keeps the book mine. If you want the step-by-step of the KDP disclosure flow itself, I broke it down in Amazon KDP's AI disclosure rules, and if you want the revision workflow, see how to use AI to revise without letting it write a word.

The edge cases that actually trip people up

The clean cases are easy. It is the gray ones that fill the forums, so let me walk the three that come up most in my inbox.

The first is heavy rewriting. A writer generates a paragraph, then rewrites it so thoroughly that barely a phrase survives. Is that still generated? The honest answer is that if the idea, structure, and most of the language originated from the tool, treat it as generated and disclose it, because the origin was the machine. The safe habit is the reverse workflow: write your own rough version first, then ask AI to critique it, so the words are provably yours from the start.

The second is dictation and cleanup tools. Speaking your novel and having software transcribe and lightly punctuate it is not generation, because you produced every word. The tool is a typist, not an author. The same goes for spell-check, grammar suggestions, and formatting help. None of that crosses the line.

The third is translation, and it surprises people. If you machine-translate your English novel into Spanish, Amazon treats the Spanish edition as AI-generated in the translation category, and you disclose it for that edition. A human translator using AI to speed up their own work sits back on the assisted side. The test is identical to prose: who produced the final words. I keep a one-line rule taped above my desk for all of this. If I cannot point to the sentence and say I wrote that, it is generated, and I tick the box.

Frequently asked questions

Do I have to disclose that I used AI to edit my grammar? No. Using AI to catch grammar problems, tighten sentences, or suggest edits on prose you wrote is assistance, and Amazon does not ask you to disclose it. Disclosure applies when AI produced the actual text, images, or translation, not when it helped you polish your own words.

Will disclosing AI-generated content hurt my sales or ranking? Amazon has stated the disclosure is internal and does not appear on your product page or affect royalties or ranking. The real risk runs the other way: failing to disclose generated content when required violates KDP's terms and can get the book removed, so honest disclosure is the safer commercial choice.

Is using AI to outline my novel considered AI-generated writing? No. Outlining, brainstorming, and planning with AI are treated as assistance because you still write the prose. The generated label attaches to the creative output itself, so an AI-built outline that you then write from yourself does not trigger a disclosure requirement.