You wrote a book with AI, now what? Disclosure, publishing, and getting it print-ready

By Tracy Atkins, founder of BookDesigner.ai and BookDesignTemplates.com · 15+ years of print files, 70,000+ authors · Updated September 13, 2026

What KDP actually asks you to declare

Somewhere in the KDP publishing flow there is a content-declaration step. Compliance trackers report the checkbox text as something like: "This title contains AI-generated content. This includes text (e.g., chapters, sections), images (e.g., illustrations, diagrams, cover art), or translations that were produced using AI tools." That wording comes from a third-party report of Amazon's copy, not a direct quote pulled from KDP's own help page on the date you are reading this, so treat it as close but not gospel. Before you publish, log into your KDP account, go through title setup, and read the live checkbox text yourself. Screenshot it and note the date. This is one of the few places in self-publishing where the exact wording matters more than the gist.

The distinction that matters is simple once you see it: AI-generated content is material an AI tool actually produced, text, images, or a translation, even if you subsequently revised it. AI-assisted content is everything upstream of that: using AI to brainstorm plot ideas, generate a working outline, or run grammar and spell check, while you write the final published words yourself. The key question KDP's framing keeps coming back to is who created the final published text and images. If the AI generated an outline and you wrote every sentence of the actual book, the content is yours. If the AI generated the outline and you mostly filled in around it, or it drafted full chapters that you lightly edited, that leans toward AI-generated and toward disclosure.

Translations get their own mention: if you use AI to translate your book into another language, even with a human editing pass afterward, that is treated as AI-generated content for disclosure purposes. Plain grammar and spell-check tools do not trigger it.

Two practical notes on how the declaration behaves once you tick it: it is not shown to shoppers on your book's detail page, and it does not affect search ranking, royalties, or category eligibility. It is a compliance record, not a marketing label. Separately, if you use KDP's A+ Content generator for your product page, Amazon's own help documentation warns that its AI-generated output "may include information that wasn't in your original input" and that you are responsible for what gets published there, with non-compliant content subject to removal. That is a different tool from the book-content declaration, but worth knowing if you use it.

Other platforms: less checkbox, more discretion

IngramSpark takes a different approach. Its Catalog Integrity guidelines explicitly prohibit "content created using automated means, including but not limited to content generated using artificial intelligence or mass-produced processes," but there is no disclosure checkbox in the submission flow. Enforcement instead happens at Ingram's discretion, after the fact, and it has occasionally caught legitimate books in the net: one publishing analyst has reported his own book, a work analyzing AI in the publishing industry rather than one written by AI, was briefly pulled from Ingram's catalog under this policy before being reinstated after he pushed back. If you distribute wide through IngramSpark, keep records of your writing and editing process in case a review flags your title.

Kobo Writing Life does not ask for a disclosure either. Its terms of service state that author works "will not be used to train generative AI technologies" on Kobo's end, which is a promise about how Kobo treats your book, not a rule about how you wrote it. Kobo does use its own machine-learning models internally to screen for hate speech and other harmful content, and its content policy flags "content generated primarily by automated tools that lacks genuine human effort, quality control, or is designed to exploit trends rather than offer genuine reader value" as grounds for removal. In other words, Kobo is watching for low-effort mass output rather than requiring a declaration for a well-edited book that happened to start as an AI draft.

The honest summary across stores right now: KDP has a specific disclosure mechanism for AI-generated content, and as of recent guides, no major distributor requires disclosure of merely AI-assisted content. That said, this is exactly the kind of policy area that shifts with little warning. Verify each platform's current wording before you publish, and do not assume today's rule holds next quarter.

What the Copyright Office cares about

Separately from any store's policy, if you plan to register a U.S. copyright, the Copyright Office has its own standard. Its 2023 registration guidance holds that if "a work's traditional elements of authorship were produced by a machine, the work lacks human authorship and the Office will not register it." That does not mean an AI-assisted book cannot be registered; it means the portions genuinely produced by a machine cannot be. Applicants have an affirmative "duty to disclose the inclusion of AI-generated content in a work submitted for registration and to provide a brief explanation of the human author's contributions to the work" whenever that content is more than a trivial amount. If you wrote the final text yourself using AI only for outlining or research, you are on solid ground describing the work as human-authored. If large stretches of AI-generated prose remain in the final manuscript, disclose them and describe your own contribution honestly on the application.

The production path, once the words are settled

Whatever the disclosure situation, most AI-assisted drafts need the same cleanup before they are book-shaped:

• Strip chat artifacts. Stray phrases like "certainly, here is your revised chapter" or inconsistent bullet formatting from a chat interface need to come out entirely; readers should never see the seams.
• Normalize headings and structure. AI drafts often shift heading styles, chapter-numbering conventions, or scene-break symbols partway through. Pick one system for front matter, chapter openers, and scene breaks and apply it consistently across the whole manuscript.
• Get a real edit pass. AI-generated prose can repeat sentence structures, flatten voice, or introduce factual errors it presents confidently. A developmental or copy edit from a human reader catches what the tool cannot see about its own output.
• Then format to spec. Margins, gutters, embedded fonts, and trim size all follow the same printer rules regardless of how the words were written. A manuscript that reads cleanly but is not built to a printer's file specification will bounce the same way any other file would.

None of this changes because AI was involved somewhere upstream. The book still has to read well and the file still has to meet the printer's specifications. What changes is the paperwork: know which disclosure box applies, fill it out honestly, and keep a record of what you wrote versus what a tool produced, in case a platform or the Copyright Office asks later.

Policy section verified via web search on August 11, 2026; re-check exact current wording against kdp.amazon.com, IngramSpark's Catalog Integrity page, and Kobo Writing Life's help center before publishing, as these policies change frequently. Related: The self-publishing checklist · Copyright page template · KDP Select vs. going wide

Frequently asked questions

Do I have to disclose that I used AI to write my book?

On Amazon KDP, yes, if the AI generated the actual text, images, or translations, even after you edited them; this is checked in the publishing flow's content declaration step. If you used AI only to brainstorm or outline and wrote the final words yourself, that is generally treated as AI-assisted, which does not currently require disclosure. Policies vary by platform and are changing quickly, so verify the live wording before you publish.

What is the difference between AI-generated and AI-assisted content?

AI-generated means an AI tool produced the actual text, images, or other creative material that ends up in your book, even if you revised it afterward. AI-assisted means you used AI for brainstorming, outlining, or tools like grammar and spell check, but you wrote the final published words yourself. KDP's disclosure requirement targets the generated category, not the assisted one.

Can I register a copyright for a book I wrote with AI help?

The U.S. Copyright Office will not register the parts of a work where a machine, not a human, produced the traditional elements of authorship. If your manuscript contains more than a small amount of AI-generated material, you have a duty to disclose that and briefly explain your own human contribution when you file. Purely AI-assisted writing, where you wrote the final text, is treated as human-authored.

Will IngramSpark or Kobo reject a book because it used AI?

IngramSpark's Catalog Integrity guidelines prohibit content created by automated means and enforce this at their discretion, without a checkbox-style disclosure step; there have been cases of legitimate books getting caught up in review before being reinstated. Kobo does not require disclosure either but reserves the right to remove low-effort automated content, and its terms state that author works will not be used to train generative AI on Kobo's side.