A self-publisher tests authorship by who wrote the first draft

The process wall

Self-publishing on Amazon's Kindle Direct Publishing runs through a fixed upload workflow: enter metadata, upload a manuscript and cover, set rights and pricing. Amazon's own content guidelines, as retrieved on 16 September 2026, insert a disclosure step into that workflow: authors must ‘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.’ The wall is not the writing itself but the point of submission, where a publisher who never reads the manuscript still requires a declaration about how it was made.

What the documents show

The guidelines draw a specific line between two categories. ‘AI-generated’ is defined as text, images or translations ‘created by an AI-based tool,’ and the guidance is explicit that this applies ‘even if you applied substantial edits afterwards’ — the test is who produced the first version, not how much was changed later. ‘AI-assisted’ content, by contrast, is content the author created and then had a tool ‘edit, refine, error-check, or otherwise improve,’ including brainstorming that did not produce the final wording; the guidelines state plainly that ‘it is not necessary to inform us’ of that category. A separate publishing help page sets a general title-creation limit — ‘10 per book format each week’ — that the page does not itself frame as an AI-specific measure, so it should not be read as part of the disclosure policy.

Authorship and limitation

The mechanism the guidelines name is about the first draft, not the final one: origin of the actual content decides the category, and edits after the fact do not move a work out of ‘AI-generated.’ This is a platform's own compliance test, not a copyright or authorship standard — the guidelines say nothing about who owns the resulting work, only what a seller must tell the platform before listing it. The same guidelines also fold AI content into an existing ‘poor customer experience’ clause, which bars content that is ‘typically disappointing to customers,’ giving Amazon separate grounds to act on quality regardless of the disclosure question.

What survives

An author can apply KDP's own test chapter by chapter: did a tool produce the actual wording, or did the author write it and use a tool only to edit, check, or brainstorm around it? That test, not a subjective sense of how much work went in, decides disclosure. Editorially, a writer publishing at volume should also note the separate weekly title-creation cap, since it constrains output regardless of any AI involvement at all.

• Apply KDP's own origin test — who produced the actual content — to each element of a book before deciding whether to disclose.
• Remember that cover and interior images fall under the same AI-generated disclosure rule as text.
• Track the separate weekly title-creation limit if publishing several books at once, since it applies independent of AI use.

The guidelines answer a narrow question — what must be disclosed to the platform — and leave broader questions of ownership, quality and reader disclosure to the author and to other rules entirely.