The Scarlet Letter of AI-Assisted Writing

Book marketing has eaten my life lately, so the LoRA Blog has been quiet. But a conversation in an AI writing group pulled me back.

An author said she was worried that using Claude while developing a new series could shut her out of book fairs and other events she has attended for years. Her earlier books were written without AI. Her new work is still her story, her characters, her decisions and her responsibility. Yet a blanket “no AI” policy could treat her as though she pressed a button and claimed a machine-generated manuscript as her own.

That is the problem. We are not having one argument about AI writing. We are having ten different arguments while pretending they are all the same.

The Scarlet Letter

Barnes & Noble CEO James Daunt recently said, “As long as an AI-written book says it’s an AI-written book … and the customer wants to buy it, then we will stock them.”

After the backlash, Daunt clarified that Barnes & Noble does not knowingly stock AI-generated books and takes steps to exclude them. His original statement was more open than the clarification, but both versions leave the same question unanswered:

What exactly is an “AI-written book”?

Is it a novel generated from a prompt and published with little human revision? Is it a human-written manuscript containing a few generated passages? Does it include a writer who uses AI to brainstorm, test a character’s motivation, critique a chapter, find continuity problems or polish a sentence?

Those are not the same creative process, but the phrase “AI-written” throws them into one bucket. Then the author is expected to disclose that label publicly, as though AI involvement is a warning readers must see before they can judge the work.

That is not transparency. It is a scarlet letter.

I am not arguing that someone should generate three hundred pages, barely read them, slap a name on the cover and demand to be treated exactly like a novelist who spent years building a book. I am arguing that “AI touched this project” tells us almost nothing about who actually authored it.

The Work Was Good Until the Process Became the Story

The panic becomes especially revealing when AI-generated or AI-assisted work passes human judgment.

A predominantly AI-generated Japanese web novel won both the Grand Prize and the Readers’ Choice Award in the 18th AlphaPolis Fantasy Novel Awards. It was selected for print publication and a manga adaptation. Those plans were later canceled after AlphaPolis updated its rules to exclude AI-generated entries.

The judges had already judged it. The readers had already chosen it. The work did not become worse after its creative process became public. The label changed, and suddenly the label mattered more than the response the work had already earned.

That case does not prove every AI-generated novel is good. Most books written by humans are not good either. It does prove that human readers can enjoy a work, and human judges can reward it, without knowing or caring which tools helped produce it.

The Witch Hunt Is Already Here

The other side of the panic is even uglier: people now believe they can identify AI writing by instinct.

They point to an em dash, a list of three, a balanced sentence, a particular transition, a polished metaphor or a word they have personally seen ChatGPT use. Then they announce that the writer is a fraud.

But AI learned those patterns from e

xamples of good human writing, and as a result, good writing suddenly became suspicious because the AI models ended up writing better than most people. 

Beyond individuals on social media accusing writers of using AI, publishing executives are warning about it too.

At the 2026 U.S. Book Show, Sourcebooks CEO and publisher Dominique Raccah said, “I fear for the industry that in this witch hunt moment, we’re going to get kept from having a much more innovation-forward endeavor.”

Sourcebooks had tested multiple AI-detection programs and found them consistently unreliable. On the same panel, Hachette Book Group CEO David Shelley said the company had chosen not to scan every submission because it did not want to create a culture in which publishers automatically distrust their authors.

Then came the 2026 Commonwealth Short Story Prize controversy.

Several regional winners were publicly accused of using AI. Readers and commentators cited supposed “AI tells,” while detection tools offered confident verdicts. The Commonwealth Foundation spent a month investigating. It interviewed the writers and reviewed working drafts, time-stamped documents, notes and other evidence of their creative process.

The Foundation concluded that AI had not been used to write the winning stories.

One of the accused stories, Jamir Nazir’s “The Serpent in the Grove,” then went on to win the overall Commonwealth Short Story Prize.

That should give everyone pause. Everyone believes they have a radar on AI so the accusations sound confident but a review of the evidence proved it was not.

A false accusation can stain an author’s reputation even after an investigation clears them. Once people decide they can smell AI in every polished paragraph, ordinary craft becomes evidence of guilt. Writers may begin roughening their prose or avoiding perfectly legitimate stylistic choices simply to appear more human.

That does not protect literature. It punishes writers for writing well.

Define the Line Before You Enforce It

Every contest, publisher, retailer and book event has the right to establish rules. But “no AI” is not a useful rule unless the organization defines what it means.

Are they banning fully generated manuscripts? Generated passages that remain in the finished book? AI-created drafts that the writer later rewrites? Brainstorming? Outlining? Research assistance? Developmental critique? Grammar correction? Accessibility tools? Translation?

A policy can reasonably draw a line. It cannot reasonably pretend there is no difference between those uses.

Writers also have a responsibility. Be truthful when a contest, publisher or contract asks how AI was used. Do not claim work you did not meaningfully direct, evaluate, or revise. Do not use AI to imitate a living writer or to conceal plagiarism. Put your name on work only when you are willing to accept responsibility for every page.

But honesty should not require self-branding every book with a warning label simply because AI was one tool in the process.

The meaningful questions are not “Did AI touch this?” or “Can I find a sentence that sounds like ChatGPT?”

Who conceived the story? Who made the creative decisions? Who accepted and rejected suggestions? Who revised the language? Who shaped the final work? Who stands behind it?

That is authorship.

Eventually, I believe most readers will stop caring about a writer’s complete tool list. They will care whether the book is good, whether the author was honest when disclosure genuinely mattered, and whether the finished work is worth their time.

Until then, we should resist both deception and purity tests.

An author who uses AI does not automatically surrender authorship. An event that cannot distinguish AI assistance from AI authorship is not protecting literature.

It is enforcing a scarlet letter.