Three months ago, Hachette canceled the US release of Shy Girl by Mia Ballard and withdrew the UK edition after weeks of online speculation about the novel’s origins.

The use of AI in the publishing industry continues to generate significant controversy.

The Shy Girl mess is not really exposing a reader problem.

It is exposing an industry identity problem.

I know that will make some people uncomfortable. Fair enough. This is not an argument that AI is harmless, or that publishing has nothing to fear from it. If anything, this shift may prove deeply destabilizing. But I still think we are asking the wrong first question.

I keep watching the reaction and thinking we are starting in the wrong place. Hachette pulled Mia Ballard’s Shy Girl after allegations of AI-assisted authorship. The book had first found traction as a self-published title, building thousands of Goodreads ratings before a traditional house stepped in, only for the deal to collapse once the production story became the headline.

And that is exactly the question nobody seems comfortable sitting with for very long.

Why acquire the book in the first place?

Probably because it had momentum. Because readers were already showing up. Because the market had signaled that something in that book was working. That is not scandalous. Publishers do that every day. They notice traction, reduce risk, and move. The uncomfortable part is that the market signal looked attractive right up until the workflow story became awkward.

That is where I part ways with a lot of the current outrage.

I think publishing should care far less about whether AI touched the process and far more about whether the final product actually served the reader.

Because publishing, at least to me, is very simple:

Publishing is about delivering the right content, to the right recipient, at the right time, through the right format.

That is the job.

Not defending some romantic mythology about how the thing was made. Not running around with a magnifying glass trying to detect whether a machine touched paragraph three. Not pretending the reader’s first emotional question is, “Wait, but what was the tool stack?”

For most readers, the real question is much more brutal.

Was it worth my time?

Did it move me, entertain me, help me, teach me, distract me, comfort me, or challenge me?

If the answer is yes, then I think we should at least be allowed to ask an uncomfortable follow-up:

Why exactly are we so obsessed with whether AI helped make it?

Now, I’m not saying all AI use is good. Obviously not.

If someone uses AI to crank out landfill in paperback form, that is obviously a problem. And no, I do not think the market magically fixes everything overnight. Bad content can still clog discovery, waste attention, and create a lot of noise. But I do think readers remain a better judge of value than purity debates. They may get fooled once. They may get fooled twice. But rubbish eventually has a smell.

And publishing does not exactly have a shortage problem right now. Publishers Weekly reported that U.S. book output topped four million in 2025, up 32.5% year over year, with the surge driven heavily by self-publishing. That alone should kill the idea that publishing needs more volume. What it needs is better filtering, better discovery, and better matches between content and audience.

And if publishing gets this wrong, the result will not just be more noise. It will be weaker discovery, weaker trust, and a weaker ecosystem.

That is why the Coral Hart example makes people twitch. Her official site markets a toolkit built by an author who says she published “200+ books with AI in one year.” That may be a fascinating experiment in speed, but speed is not the same thing as value. Volume is not the same thing as relevance. And scale is not the same thing as craft.

Still, even that example proves something important.

The real issue is not “AI or no AI.” The real issue is whether the output is any good.

That is the line I care about.

If AI helps an author sharpen structure, test angles, improve pacing, adapt content for new formats, or economically serve a niche audience that would otherwise never get that content at all, why should publishing treat that like contamination?

We do not disclose spellcheck. We do not disclose every editorial intervention. We do not disclose every ghost in the machinery of making a modern book.

Yes, AI is different. I get that. But the reaction still feels wildly selective. Sometimes it sounds less like principle and more like vibe.

And that is why I find some of the public responses slightly theatrical.

The book business loves to say it is reader-first. Fine. Then be reader-first.

If many readers genuinely do not care whether AI had a hand in the workflow, should a publisher care?

My answer is: only if it affects the publisher’s brand promise, legal risk, or commercial trust with its ecosystem. That is a business decision. It is not a universal moral commandment handed down from Mount Goodreads.

And I do think there is a difference between hidden slop and real creative experimentation.

Stephen Marche’s Death of an Author, released under the hybrid pen name Aidan Marchine, was explicitly framed by Pushkin as a human-machine project, with Marche saying the novella was 95% written by AI and 5% by him through outline and prompting. That is not a secret process. It is the concept itself. You can love it, hate it, or roll your eyes at it. But it forces the right question: if the reading experience works, what exactly are we objecting to?

For me, this is where the whole debate gets stuck in 2024 while the world has already moved on.

The relationship between humans and AI is still immature. The boundaries are moving. The quality is improving. Detection is messy. Even the industry’s response shows that we are dealing with a fluid situation, not a settled one.  The Authors Guild’s “Human Authored” certification still allows limited AI use, including for brainstorming, outlining, research, and auxiliary materials. The UK Society of Authors has launched a similar “Human Authored” scheme. In other words, even the people drawing hard lines are quietly admitting the line is not actually hard.

That is why I think the conversation should move away from purity and toward performance.

Did the content find its people? Did it deliver value? Did it earn attention? Did it deserve to exist?

And yes, format matters too.

Because publishing is not only about bound paper sitting on a shelf. For one person, the right format is a hardcover on the couch. For someone else, it is an audiobook in the car. For someone else, it is a short video that leads them into a longer story later. For someone in an airport, the right format may be totally different from what works for them on a quiet Sunday at home.

If the mission is to deliver the right content, to the right recipient, at the right time, through the right format, then AI is not automatically the threat.

Irrelevance is the threat. Noise is the threat. Bad judgment is the threat. Flooding the market with forgettable sludge is the threat.

So no, I do not think disclosure should be treated like some sacred rule in every case. If someone wants to mark AI-written parts, fine. If someone wants to be transparent because that is part of the artistic concept, great. If someone does not disclose because the output stands on its own and serves the reader well, I am not convinced that is automatically a sin.

My bar is simpler.

I care far more about quality than I do about mandatory disclosure.

That will annoy people. Especially people whose definition of authorship is tied very tightly to process. I understand that. Authors have every reason to care. Their identity, scarcity, and status are bound up in this shift. And they may be right to worry, because this is not a cosmetic change. But publishing cannot become a museum of approved workflows. It still has to function in the real world.

And in the real world, the winner is usually not the loudest defender of purity. It is the person who best matches content to the audience.

That is why I think Shy Girl is less a scandal about one book and more a stress test for the whole industry.

What does it mean in 2026 to be a publisher?

What does it mean to be an author?

And maybe the sharpest question of all:

Do we care more about how the content was made, or whether it truly earned the reader?

Because if publishing keeps arguing from purity alone, the real decisions may get made without it.

That is the conversation worth having.

Sources:

• The Guardian: Hachette pulls horror novel Shy Girl after suspected AI use

• Publishers Weekly: Book Output Topped Four Million in 2025

• Coral Hart AI: official site

• Pushkin: Stephen Marche and Death of an Author announcement

• Pushkin preview: Death of an Author described as 95% written by AI and 5% by Stephen Marche

• Ahrefs: “AI Content Wasn’t Good Enough. Now It Is.”

• https://authorsguild.org/human-authored/faq/

• https://humanauthored.co.uk/faqs/

• https://societyofauthors.org/2026/03/11/human-authored-scheme-launched-as-generative-ai-threatens-authors-livelihoods/