Under US Copyright Office guidance, material produced by a machine with no human authorship is not copyrightable — the Office states plainly that a work whose traditional elements of authorship were produced by a machine "lacks human authorship" and won't be registered. A human selecting or arranging AI-generated material in a sufficiently creative way can make the resulting arrangement itself protectable, even though the underlying AI-generated content is not. This is not legal advice.
What the guidance actually says
| Situation | Copyrightable? |
|---|---|
| Image produced entirely by an AI tool, no human creative input | No — lacks human authorship |
| A human selects or arranges AI-generated images in a sufficiently creative way | The selection/arrangement can be protected; the underlying AI content is not |
| A human substantially edits or adds original elements to an AI-generated image | The human's contribution can be protected on a case-by-case basis |
This table reflects the US Copyright Office's published registration guidance, not a CrayonMill legal opinion.
Why the Office draws the line at authorship
Copyright has always protected human creative expression, not the output of a process alone. The Copyright Office's 2023 guidance applies that same principle to generative AI: if the traditional elements of authorship in a work — the expressive choices that make it what it is — were produced by a machine rather than conceived by a person, the work doesn't meet copyright's authorship requirement. A prompt alone, however detailed, does not turn the person who wrote it into the author of everything the AI then produces, because the guidance ties authorship to who actually determined the expressive content, not who requested it.
What "selection or arrangement" protection actually covers
The guidance draws a specific distinction: a human choosing which AI outputs to include, in what order, combined with what other elements, in a way that is itself creative, can make that overall selection or arrangement an original work of authorship — separately from whether the individual AI-generated pieces inside it are protected. In practice, that means the compiled, arranged book (a specific sequence of pages, a specific combination of theme and layout choices a person made) sits on different footing than any single AI-drawn image considered alone.
What KDP requires regardless of the copyright question
Separately from copyright registration, KDP's own content guidelines require publishers to disclose AI-generated content — text, images, or translations — when publishing or republishing a book, and state that the publisher "is responsible for verifying that all AI-generated and/or AI-assisted content adheres to all content guidelines, including by complying with all applicable intellectual property rights." That disclosure obligation applies whether or not the underlying images end up being copyrightable, and it sits on the author, not on the tool that generated the images.
Why this doesn't answer "can I sell it"
Copyrightability and the right to publish and sell a book are two different questions. A page can be entirely uncopyrightable and still be perfectly legal to include in a book you sell, provided the generator's own terms allow commercial use and the content doesn't infringe someone else's existing copyright. The reverse also holds: copyright protection on an arrangement doesn't by itself grant a commercial license from the tool that produced the underlying images. See the commercial-use guide for the question that actually controls whether a given book can be sold.
Not legal advice
This page summarizes published US Copyright Office guidance as a starting point for understanding the general shape of the issue. It is not legal advice, and it doesn't address every fact pattern (co-authorship, substantial editing, jurisdictions outside the US). An author with a specific, high-stakes question should talk to a qualified attorney rather than rely on a summary page.
Sources
Last verified 2026-09-04. Updated 2026-09-04.