Barnes and Noble Press asks its own AI disclosure question

The writer's problem

An author publishing the same book across several print-on-demand and ebook retailers quickly discovers that disclosing AI use does not mean the same thing twice. Barnes and Noble Press, the company's self-publishing arm, asks authors during its Build-a-Book process to confirm whether generative AI was used in a project's cover, manuscript, images or translations, using its own help article to define the categories.

What the documents show

The article states that any element made with AI must then be marked 'AI-Assisted or AI-Created,' and defines 'AI-Created content' as anything 'created using Generative AI,' adding it 'would still be considered AI-Created if you used AI to generate the content and edit it afterward.' 'AI-Assisted content' is content the author created and then used AI 'to edit it or expand on it.' Completing this AI Disclosure, under the project's Rights and Other Information tab, is required before a project can be published. Those category names differ from Amazon KDP's own Content Guidelines, which use 'AI-generated' and 'AI-assisted' with a similarly worded but separate test, and from Apple's Books asset guide, which asks for an AI Generated by artist-role credit rather than a yes-or-no declaration field. All three require disclosure of substantially AI-made content and exempt lighter editing assistance, but none uses the same label or submission mechanism.

The editorial choice

A publisher distributing wide has to translate one underlying fact, which parts of this book were AI-made, into three different vocabularies and forms. This is an editorial practice, not settled by any retailer's page: keep one internal record of what was AI-Created versus AI-Assisted in plain language, then map that record onto each platform's specific category at submission time, rather than answering each retailer's question from memory.

What stays with the author

B&N Press's article does not resolve what happens if an author disagrees with how a tool's contribution should be classified, and it keeps the author responsible for the separate Content Policy regardless of which AI category applies, including that the project 'does not infringe on or violate intellectual property rights or copyright.' Reconciling three retailers' definitions for one manuscript remains work only the author or their publisher can do.

• Would this book's AI-touched content count as AI-Created at B&N Press, AI-generated at KDP, and require an AI Generated by credit at Apple, or fall under the lighter-touch category everywhere?
• Has the AI Disclosure field actually been completed, since B&N Press ties it to the ability to publish?
• Does the content still meet each retailer's separate intellectual-property requirements, independent of the AI label?

The mismatch between these categories is not a flaw in any one retailer's policy; it reflects that no shared industry standard yet defines what counts as AI-made for a self-published book.