What Publishers Must Ask Themselves Now

Good morning. I’m on my second Americano, and it’s not even 9 am. Today is the last day I will be 53. My birthday is tomorrow (Friday), and I can hardly believe I’m turning 54. The past year, I dealt with a lot of health stuff, which I’m hopefully leaving behind for a while. The good news is that I’ve lost close to 80 pounds since my gastric sleeve surgery last December, and I’ll know what my bloodwork looks like on Monday. We are moving my daughter into college in Rhode Island on September 5th, so it’s an emotional time. She is my anchor, and I’m sad she’s leaving, but also so excited for her. My son (22) is now living back at home (he graduated from college in May), which is nice. Our compromise at night is that he’ll sit through an hour of whatever I’m watching and then I let him turn on sports. It works out! And finally, I start teaching college again on August 25. I’m looking forward to it.

Today’s newsletter explores the AI conundrum publishers are finding themselves in. It’s worthy of a discussion.

Now that we’ve all said as much as we could about the Jerry Falade book and subsequent fallout, something has been gnawing at me: If a publisher has an author who has been accused of possible AI use in their writing, will they always cancel the book? The answer is worth exploring.

Let’s start with Daggermouth by H.M. Wolf, which Scarlett Press (Simon & Schuster) published on July 28. This is the book The Atlantic used as an example in its July 27 article, "Is This What Comes After AI Slop?" The Atlantic framed it in the context of the very faulty study “Generative AI floods and dilutes the market for books,” which is not peer-reviewed and was just released (Sidenote: The lead researcher on the study, Tuhin Chakrabary, has been finding me on every platform, including this one, to belittle me because I am critical of the study. He has been blocked.). The researchers uploaded Daggermouth, which was initially self-published, to Pangram, which said it was 60% written with AI. Additionally, there are some identical sentences in Daggermouth that appear in other authors’ books in the genre. In my opinion, it is difficult to unequivocally prove that AI wrote the sentences because so much genre fiction uses the same language. If you don’t read genre fiction like romance and romantasy, you wouldn’t know that.

Scarlett Press and Simon & Schuster stood by their author and went ahead with Daggermouth’s publication. They also acquired more books from the author. In its first week, Daggermouth sold over 31,000 copies (Bookscan) and debuted at #1 on The New York Times Hardcover Fiction Best Seller list. The people have spoken.

Then there is Shy Girl by Mia Ballard, which was also initially self-published. Hachette acquired it and was set to publish it in May, but a series of events resulted in them canceling the book. I want to be clear here and say that I was never comfortable with the way this went down, and my instinct was correct. Last night, I came across a piece by The Drey Dossier that laid out what happened. I urge you to read it:

First, I was never comfortable with Hachette canceling a book that Pangram deemed 78% AI-written. Secondly, the way this manuscript was obtained was unethical. Third, you need to connect the dots between Pangram’s founder and publishing consultant Thad McIlroy. From The Drey Dossier:

He writes that he learned about Shy Girl at the beginning of February from Asia Laird, whom he describes as recently hired at Pangram. When I read that name, I wanted to know who she was, so I looked her up, and what I found reframed the entire story. Her LinkedIn says Sales and Business Development at Pangram. ZoomInfo lists her as Founding Account Executive. She is one of Pangram’s first hires, and her job is to sell their product, and their product is the AI detection tool whose scan produced the 78 percent number that the entire world has been repeating ever since.

So the person who surfaced this story to the man who brought it to the New York Times was a sales employee at the company whose product became the key evidence. That is where the whole thing starts.

McIlroy also writes that he’s known Pangram’s CEO, Max Spero, “for a couple of years” and has “a lot of faith in the seriousness of his technical approach,” which matters because his consulting firm works with publishers and with the tech companies that sell products to publishers, meaning this was a pre-existing professional relationship between a detection company and a consultant whose entire business depends on exactly these kinds of tools.

And here is what makes the timing so interesting. On January 23rd, weeks before Asia Laird brought the story to McIlroy, Max Spero had already posted Pangram’s scan results publicly on X: 78 percent AI detected. And right there in the preview text of the tweet, visible without clicking anything, you can see OceanofPDF.com, a well-known book piracy site.

This is troubling, to say the least. I already knew some of the story, which is why I wasn’t comfortable with how it was portrayed in The New York Times, and felt like Hachette rushed to judgment. It didn’t help that someone made a 3-hour YouTube video dissecting the book to “prove” it was written with AI. All that said, is pirating a book to run it through Pangram an ethical way to discern whether it was written with AI? Two wrongs don’t make a right, as the saying goes. It is also hard to swallow that the person who brought this story to the Times and was quoted in the story had a professional connection to Pangram. None of this was disclosed by The New York Times. I have to wonder whether they were told about how the book was obtained and the relationships involved. Hachette claimed they did a “thorough and lengthy review of the text” after the Times story broke. How one can do such a thorough review of a text in 24 hours boggles the mind. So does the fact that the book had already been published in the UK (they discontinued it). So, you’re telling me that the book was acquired and no one edited it? Someone at Hachette had to read it, right? (Stick with me here; I am getting to my point)

Next, there’s Jerry Falade. We’ve talked ad nauseam about his book Call Me I’ll Hide the Body (great title), its 14-way auction, the $2.4M advance, and how it all fell apart. According to The Wall Street Journal, Falade has retained legal counsel. This is about to get very interesting! Still, the question remains: How did a manuscript get through 14 editors if it allegedly had “hallmarks of AI?” Anecdotally, I’ve spoken to people who read it, and they uniformly loved it. I still think Falade should self-publish the book immediately.

All of this leads me to a question publishers must ask themselves: How much AI are they willing to live with? If Scarlett Press moved ahead with Daggermouth’s publication and acquired even more books from the author, they are clearly comfortable with a 60% Pangram score. I would be, too, because Pangram is imperfect. I’ve mentioned before that I’ve uploaded my own writing, and it was scored at 80% AI-written. If 60% is acceptable to Scarlett Press, why was 78% unacceptable to Hachette in the case of Shy Girl? There is no doubt in my mind that the book, had it been traditionally published, would have sold well. And there is the crux of the situation: Not all books are created equal. Just because a book is traditionally published, it doesn’t mean it is a literary masterpiece.

What I find faulty in the AI-diluting-the-market-for-books theory is this: The market has been diluted with what a lot of us would consider mediocre books for a LONG time. The masses do not read literary fiction. Most of the time, we are lucky that people buy books and read them, period. My philosophy is that publishers must diversify their lists because you really do need to have something for everyone. If we look at fiction bestseller lists, they are populated with very commercial fiction. Commercial fiction, in the most basic terms, is more about the story than the writing. Compare it to, say, a buddy cop movie from a big studio versus a film from a studio like A24. It’s pop music versus classic jazz. It’s Costco versus a boutique.

When an agent or publisher decides to pull a book due to alleged AI use, they may think they’re doing the ethical thing, but unless you can 100% prove it, isn’t it a bit unethical to cancel a book that was obtained illegally and uploaded to an AI detection platform without the author’s permission? Isn’t it unethical to give The New York Times information about a book you only found out about because of a relationship you had with Pangram? Speaking of Pangram, they are the only winners in this situation. They are getting so much free PR from these stories that it helps their funding rounds. They just raised $9M. They are making money from these accusations.

Some things I know for certain: If Shy Girl wasn’t called out, Hachette would have gone through with its publication. Scarlett Press was not going to pull Daggermouth over a 60% Pangram score. Had Jerry Falade’s deal gone through, Minotaur would have published Call Me I’ll Hide the Body, and it would have been optioned for film/TV. What does this tell us? It tells us that consumers drive the market, and the market kind of doesn’t care if something is written with AI. I’m not talking about literary fiction readers or other authors. I am specifically referencing the masses. You know, the ones who will stand in line for new Starbucks merch. It’s not that it’s necessarily good; it’s that it is there. In the case of movies, it is more about being entertained than getting too deep. We’ve been living in precarious times for a decade. The horrors persist, and the masses want to give their brains a break.

I am by no means saying that publishers should go hog wild and ignore AI for the sake of sales. I am saying that there are cases where they will do just that. I feel for debut authors because they will be on the receiving end of the most scrutiny. It is much easier to cancel a book from an author with no sales track than it is to cancel a very successful self-published book that has sold phenomenally well.

Publishers must revisit what is ethically acceptable to them. They can put the onus on authors to disclose AI use, but they must also decide if they are okay with people obtaining manuscripts and uploading them to an AI detection platform that is directly or indirectly profiting from the industry. You can slice and dice it any way you want, but this is the truth. There will always be bad actors who produce an obscene number of books using AI and upload them to Amazon. That problem won’t be solved any time soon. My concern is that publishers notoriously say one thing but do the opposite. I’ve been in too many meetings where the phrase “we’ll just let this go and see what happens” is spoken to believe that every book suspected to be written with AI is being canceled. The books that have the best shot at remaining on a publisher’s list are those with a lot to lose if pulled. We have seen it with questionable memoirs, and we will see it with possible AI use. Every Big Five publisher (I’m focused on them because indie presses are far more discerning) must ask themselves what they’re willing to live with because what’s happening now isn’t working. They must do better.

END NOTES:

What I’m Reading: Offseason by Avigayl Sharp, which is longlisted for The Center for Fiction 2026 First Novel Prize. I’ve been reading literary debuts from the longlist, and they have been terrific (I’ve also read Like an Animal and Whidbey). I’m only 60 pages into Offseason. It’s about a young woman who takes a job teaching at an all-girls boarding school on the Eastern Seaboard and tries to figure out where her life went wrong.

What I’m Watching: I binged Amazon Prime’s new YA series Sterling Point, and I am already rewatching it because it is so good. There’s an innocence to it: Summer on a lake where there is really no cell service, kids who (mostly) hang out in person and talk to each other, a romance, and a family secret. I also just finished watching Rooster on HBO, and absolutely loved it! Steve Carrel was born to play that role.

What I’m Listening to: Robby Hoffman on the Smartless podcast. Great interview.