In May 2026, James Daunt was asked what Barnes & Noble would do with books written by artificial intelligence.
His first answer was disarmingly simple.
He had no problem selling a book, he told Jenna Bush Hager on Today, as long as it did not masquerade as something it was not.
If an AI-written book said what it was, did not appropriate someone else’s work, and customers wanted it, Barnes & Noble could stock it.
The answer placed a great deal of weight on the label.
Tell the reader what the object is.
Let the reader decide.
Then people objected.
Within days, Barnes & Noble clarified its position.
The company did not endorse or intentionally sell AI-generated books, Daunt said. It took active measures to exclude them from its online offerings and did not knowingly order them for its stores.
Barnes & Noble also demanded that publishers label AI-generated books.
In the space of a few days, the same institution had produced two answers.
AI-generated books were sellable in principle.
AI-generated books were actively excluded in practice.
That is not necessarily a contradiction.
It may be what happens when the shelf is asked a question it was not built to answer.
A shelf looks passive.
It is not.
Before a book reaches it, someone has already decided that the object is a book, decided who wrote it, assigned it an ISBN, placed it in categories, described it to distributors, and offered it for sale.
The retailer inherits all of those decisions.
Then it has one more to make.
Do we put it here?
For most books, the earlier classifications make that question nearly automatic.
An AI-generated book disturbs the sequence because one description no longer seems sufficient.
A novel can still be a novel.
A history can still be a history.
The title can still have an author.
But now another fact about the object’s production begins demanding visibility.
Who—or what—generated the language?
And who needs to know?
Amazon has already answered one version of that question.
Its Kindle Direct Publishing rules distinguish between AI-generated and AI-assisted content.
If an AI tool generated the text, images, or translations, Amazon requires the publisher to disclose that fact when submitting the book.
And the category survives editing.
If AI generated the original content, Amazon says it remains AI-generated even if the human later applies substantial edits.
But if a person creates the content and uses AI to brainstorm, refine, edit, or improve it, Amazon classifies that as AI-assisted.
That does not require disclosure.
The dividing line is therefore not simply:
Was AI involved?
It is:
Who generated the content?
And Amazon locates that distinction at the beginning.
Editing does not erase origin.
Barnes & Noble Press, the company’s self-publishing platform, draws a remarkably similar line.
It distinguishes AI-Created content from AI-Assisted content.
If generative AI created the content, later editing does not make it human-created.
If the person created the content and subsequently used AI to edit or expand it, the work is AI-Assisted.
The author must make the appropriate disclosure before publishing through the platform.
Again, the classification attaches to origin.
The human may then select.
Reject.
Rewrite.
Rearrange.
Correct.
Perhaps substantially transform.
But the original act of generation remains part of the object’s identity.
That is a different way of locating authorship from the one we encountered on Stephen Hawking’s covers.
Hawking’s communication machinery became progressively more capable.
Eventually his predictive system could supply so many characters that he physically entered fewer than 20 percent of them.
Yet the attribution remained Stephen Hawking.
The machine was understood as reducing the distance between his intention and its expression.
It was predicting Hawking from Hawking.
The cover located authorship in the person whose thought, judgment, and direction produced the work.
The contemporary self-publishing policies locate something else.
They ask who generated the material first.
Those measurements do not necessarily conflict.
But they are not the same measurement.
A human can exercise enormous judgment over AI-generated material and still have Amazon classify the material as AI-generated.
A machine can supply enormous amounts of predicted language and leave human authorship undisturbed.
One institution looks toward intellectual direction.
Another looks toward origin.
The difference matters.
It matters even more because disclosure itself is beginning to split.
Amazon KDP requires disclosure to Amazon.
Its policy does not, by itself, require the resulting book to carry an AI label for the reader.
Barnes & Noble Press requires disclosure through its publishing system.
Daunt’s public comments introduce another audience entirely.
The customer.
His position makes the label part of the transaction.
If an AI-generated book reaches the shelf, the buyer should know what it is.
Those are three different acts hiding inside one word.
Disclosure to the platform.
Classification by the publisher or self-publishing service.
Disclosure to the reader.
They answer different questions because they serve different decisions.
The platform wants to know what it is accepting.
The retailer wants to know what it is stocking.
The reader wants to know what is being offered.
This is where the shelf becomes more interesting than it first appeared.
Barnes & Noble’s two answers in May were not simply yes and no.
They exposed two different functions of the retailer.
One answer came from commerce.
If an honestly labeled product has quality and customers want it, why should a bookseller refuse to sell it?
The other came from institutional selection.
Barnes & Noble does not presently want AI-generated books filling its catalog or its stores.
It takes measures to keep them out.
The first asks:
Can this be sold?
The second asks:
Will we stock it?
The distinction existed before artificial intelligence.
Bookstores have always carried only a fraction of the books that could legally and commercially be sold.
AI simply made the distinction visible.
And that may be what these new policies are really doing.
They are not yet giving us a universal definition of authorship.
They are building instruments for institutions that need to make decisions before that definition exists.
Amazon draws a line at generation.
Barnes & Noble Press draws a similar line.
Barnes & Noble’s stores add provenance, labeling, quality, publisher reputation, and customer demand.
Different institutions.
Different audiences.
Different decisions.
The machine may have generated the language.
A human may have transformed it.
A publisher may have accepted it.
A platform may have classified it.
A retailer may still refuse to stock it.
And a reader may still want to know.
By the time the book reaches the shelf, the question is no longer merely:
Who wrote this?
The shelf has inherited that question and added another.
What do we need to call this before we can put it here?
WE&P by: EZorrillaMc&Co.
Sources
Amazon Kindle Direct Publishing — Content Guidelines
Defines AI-generated and AI-assisted content. Requires disclosure to Amazon of AI-generated text, images, or translations, including material substantially edited afterward; AI-assisted work does not require disclosure.
Barnes & Noble Press — Use of Generative AI in Your Project
Defines AI-Created and AI-Assisted content. AI-generated material remains AI-Created after subsequent editing, and an AI disclosure must be completed before publication.
Publishers Weekly — “James Daunt Looks to Clarify B&N’s Position on AI-Generated Books” (May 21, 2026)
Documents both sides of the Barnes & Noble episode: Daunt’s Today comments that transparently identified AI-written books could be stocked, and his subsequent clarification that B&N actively excludes AI-generated books while stopping short of an absolute ban where reputable publication, labeling, and clear customer demand exist.
Publishers Weekly — May 25, 2026 issue
Print-issue record for the Daunt clarification and its publication context.
Travel in Life 