"Publishing is under a full-fledged assault from AI slop right now."

WTF is going on in publishing?

Last month, I talked over g-chat with Keith Riegert, president of The Stable Book Group and CEO of Ulysses Press, VeloPress, and Perfect Bound Technologies. ‘Twas a few short weeks before Daggermouth came out amidst AI accusations and publishers killed a $2.4 million book contract over alleged AI use. The publicist’s pitch had mentioned eroding reader trust, a topic near and dear to my heart. I thought I knew where the conversation with Keith would go.

Reader, I did not.

Below, I’m sharing our wide-ranging discussion of AI’s role in the industry, why he instructed his employees to use AI “an hour a day no matter what”, those pro-AI comments from Barnes & Noble’s CEO, and why Keith believes that, when it comes to AI adoption in publishing, “it’s already over.”

Though he and I disagree on many points, I found his candor and arguments fascinating (and, at times, infuriating) and especially appreciate his take on what aspiring and current authors should do during this turbulent moment. What do you think? Let me know in the comments!

Me: Hi, Keith!

Keith: Good morning!

How are you? Thanks for chatting with me today!

Happy to be chatting with you! I’m cold. California is cold and foggy and I don’t care for it.

How are you?

I’m good, thanks! NYC is quite hot today, but nothing like the heat waves of late, so I’m trying not to complain.

I’m looking forward to getting back home to the heat!

It’ll be waiting for you!

Before we dive into all things AI-in-publishing, can you share a bit about your background and what you do as president of The Stable Book Group and CEO of Ulysses Press, VeloPress, and Perfect Bound Technologies?

Sure thing.

I oversee the strategy at the collection of companies, which are now rolling up into one publishing operation.

The idea behind the Stable is that independent publishing is extraordinarily difficult--it’s very labor intensive and margins on book sales are thin, so by joining indie publishers together, we can share in the economies of scale (think accounting, production, marketing, etc.) of a larger organization while maintaining the unique voices of smaller companies. Ulysses is my family’s company, my parents started it in 1983, and is one of the members of The Stable.

Perfect Bound is the tech that undergirds the entire operation--from managing title metadata to royalty accounting.

The idea is to automate as much of the publishing process as possible to reduce the workload.

Finally, we have a distribution operation focused on small independent publishers that we created in partnership with Hachette Book Group.

(Fun fact: My very first book, a co-written nonfiction/humor book based on a Tumblr of the same name, came out from Ulysses in 2010! It was called Stuff Hipsters Hate. “Indie-sleaze” before we had that term, lol.)

I remember that book well!!

I was an acquisitions editor

Was that Kelly that acquired it?

Or Nick?

It was Kelly! My friend and I were, like, 23 and had so much fun writing and promoting it.

Just to be crystal clear, these are independent presses in the traditional sense (operating like the Big 5 yet smaller than the Big 5)—not “indie publishing” as another term for self-publishing, correct?

Correct. Traditional publishing houses that are unaffiliated with the larger corporate publishers.

Got it, thank you. So, like all publishing companies, you must be talking about AI internally! Do you remember when it first started coming up in discussions about operations, acquisitions, etc.?

I was pretty early to AI (at least as far as publishing goes). My big oh-shit moment was about three years ago when I asked ChatGPT to write Python code that would scrape Urban Outfitters’ website and return a dataframe (spreadsheet) of all the books it was selling. I spent years learning to code in Python (very mediocrely) and that thing wrote better, cleaner code than I could ever write.

So as soon as that happened, I started paying very close attention. My guidance to our employees was to use AI for an hour a day no matter what. Use it for whatever you were doing--writing emails, drafting press releases, coming up with marketing plans. Anything that didn’t include the actual writing and/or production of books (that’s the line we don’t cross). My goal for them was to get used to what it could do, what it couldn’t do, and how quickly that was changing.

Now, with agentic AI, we are using it very heavily in the operations of the companies.

Oof! My girlfriend is a Python coder and reports that it’s made her skillset largely irrelevant.

Yup.

It’s brutal.

Are you worried about AI eroding employee’s nontechnical skills—especially those who are early in their careers, who might not have the experience to look at AI output and gauge if it’s any good or not?

Absolutely. We call it “cognitive surrender”.

One of the best examples is our sales conference script generator. This is a custom GPT that we trained on 5 years of sales conference scripts and the corresponding book information (tip sheets).

This process used to be absolutely brutal for young editors. Write, pitch, get torn to pieces, rewrite, rinse, repeat.

It took weeks. But you built the muscle over time. Now you can just drop in the tip sheet and it generates the first draft for you (probably 80% baked). Huge time savings. But that repetitive muscle memory is gone.

To be clear, I consider myself a pretty heavy user of AI. And I absolutely hate that it is here.

The loss of that muscle memory—it’s probably eroding editors’ ability to take on the last 20% of the task, right? How do you safeguard for that?

You try to balance the amount that AI is doing with what editors still write on their own. Fortunately, this is an industry that has, from the bottom up, been very wary of AI and hesitant to use it. So it’s not as big a problem in publishing as it is in other industries, I think.

I push the use of AI on our team and it’s very hard to get broad adoption.

In the end, as a leader, it comes down to how much productivity increase you are willing to give up for the good of employees’ career development. Which is an awful position to be in.

Are you also soliciting employee input about whether they feel it improves their productivity and skillset and aligns with their values? The environmental impacts as well as the AI companies’ political involvement (mass surveillance contracts, OpenAI as a top Trump donor, etc) might be part of why it’s hard to get that broad adoption.

Yes, definitely. But, at this point, if you are not AI literate, you are unemployable. So, while I fully respect anyone’s values or thoughts on the evils of AI, I would be doing a huge disservice to them to not push for adoption. You can smash the looms as much as you want; it’s already over.

And I think you would be hard pressed to find anyone who hated AI more than I do.

I have a very dark outlook on this technology. I think it’s worst thing to happen to humanity, full stop.

This is so interesting because I hear people on one (my) end saying it’s awful, and those on the other pressing employees to use it or get left behind but also, typically, extolling its virtues. But you’re in a different position! Can you say more about why you feel you couldn’t say, “I know AI is out there, I know they’re pressing it hard, but we’re not going to use it here”?

I have a background in business (Stern MBA) and, despite being a good west coast liberal, I consider myself a free market capitalist. Not because I think it’s the best approach--but because I believe that is the mean to which everything reverts.

And in that marketplace, it is a zero-sum game and the companies that adopt these technologies will be the ones left standing and the ones that don’t will die out. That’s my fear at least. And, as a CEO, my job is to protect the wellbeing of our companies and our employees at all costs.

So, I have to hold those contradicting positions on my own.

In the end, though, I don’t believe AI will work out.

The only way to recoup the investment that these companies have made is to automate the entire labor force and capture that value. And if you do that, who the fuck is left to buy any of the products you’re churning out?

IT MAKES NO SENSE!!

[End of rant]

We definitely agree on that point! To be flip, 90% unemployment is not good for...the entire economy, let alone book sales!

Yeah. Figure that one out Sam Altman.

Let’s talk about AI in the publishing industry. After the B&N CEO said he’d sell AI books in-store if consumers wanted them, your publicist sent me an interesting pitch about your take on AI in our sector. It included the term “trust recession”—can you say more about that?

Publishing is under a full-fledged assault from AI slop right now.

The number of ISBNs that came on the market between 24 and 25 jumped by one million.

That’s insane. Meanwhile, the number of “traditionally published books” rose 6.6% from 602k to 642k

Mind-boggling!

So the market got absolutely flooded by crap. And Amazon is has been the primary culprit. You can publish any AI-crafted book on KDP for free as long as you tell Amazon (not the end consumer) that the book used AI.

What we are seeing as publishers is that as soon as you have a top-performing book on Amazon, you immediately have a host of strange competitors with authors that don’t seem to have ever existed.

Yikes

Kara Swisher has a great story of half a dozen “biographies” of her popping up as soon as her book was **announced**

Just pure slop making an SEO play.

What’s worse, is that the AI books being generated are often using the book material that was pirated by AI in the first place.

I’ve had that happen to friends too—ludicrous “biographies” suddenly for sale.

It’s so depressing. And it could have a tremendously negative impact on publishing (making it more difficult to establish a book), for authors (there won’t be money to pay for books) and consumers (who have to be very careful about what they purchase, consume, believe.)

I’ve seen a number of readers saying, “I’m not going to buy any book published after 2023 because I want to be SURE the author didn’t use AI.” Which is...a really good way to tank an industry you love.

And that actually brings us full circle to why I’m pushing AI adoption at our company--we need to be able to beat those bad actors and we won’t succeed without using those same tools they are using.

I’ve heard that too. But, from what we are seeing in booksales, the majority of readers don’t care.

Don’t care if the author is human or AI, you mean?

Yes. They’ll buy the AI slop.

But it’s the slow erosion that will be really damaging.

[Sad Ben Affleck meme]

Ironically, I believe the effect will be to reintroduce the value of “the gatekeeper”.

The professional book publishing houses. Top editors. Known authors. Independent book stores with human, book-loving employees making legit recommendations.

As a publisher, I’m excited about the brewing backlash more than anything else.

This aligns with what I keep telling authors, which is “let your audience get to know you as a human” and “get to know your friendly local bookseller”. But, question—if readers don’t care if it’s slop or not, why will they gravitate toward “gatekept”/vouched-for books?

If Amazon becomes impossible to navigate in a trustworthy way, it loses a lot of market power. And that power shifts to other avenues. We’ve already seen it with the resurgence in indie bookstores and the gravitation of people to the “third place”.

And the audience, the part of the US population that are avid readers, is actually quite small. Sure, slop will continue to sell--but the arbitrage opportunity for AI slop producers starts to disappear.

What’s the point of spending 100k tokens to produce an AI book that nobody buys?

That’s the pivot point.

Yes! Oh, I so hope the backlash means we all start to value community and quality over convenience and cheapness and AI slop again.

I don’t want to take up too much of your time, but I’d love to hear: Given the period we’re in and the pivot point that’s (hopefully) coming, what’s your advice to aspiring and current authors?

One of the things we’ve always done in book publishing is look to the music industry as our big sibling--tech tends to hit it first and we watch how it adopts. Well, vinyl record sales fifteen-twenty years ago were about 2.6 million units. In 25 it hit 46 million. Vinyl is a billion dollar industry again (or for the first time?). The analog revolution is upon us.

My advice to aspiring and current writers is this: don’t fucking use AI to write your book.

Be aware that self publishing is going to get much more difficult in the years to come, especially as AI search (GEO/AEO) replaces SEO.

And understand that this market is going to be in major flux for a long time to come. You need to be aware of what is selling in the industry and what isn’t; publishers are going to be making smaller, more conservative bets.

When you say “be aware of what is selling,” do you mean in the way of writing toward trends (romantasy etc) or more like confirming you’ve got a commercial hook, etc.? It’s so tricky in this industry with its many-year lead time.

So be prepared for a publishing landscape that looks different than it did ten years ago. Right now, platform means everything to publishers (because the traditionally relied on modes of marketing have stopped working effectively). So build up your audience (and definitely let them know you as a human) before you ever pitch your book. It’s not about the size of your audience--it’s how engaged you keep them.

At the U.S. Book Show, I sat on the “Data-Driven Publishing” panel and the main point I wanted to make was “data-driven” publishing is over.

There is a problem in the industry that bridges publishers and retailers, where everyone is relying on data without really understanding what happens when you rely on data.

Goodhart’s Law is “when a measure becomes a target, it ceases to be a good measure”. Essentially, when everyone starts optimizing on the same data sets (what’s selling), the opportunity vanishes immediately. 20 publishers coming out with hockey romance saturates the market very quickly.

Great point! Chasing after the wrong thing, assuming it’s the independent variable...

So, as an author, you need to understand that dynamic in the industry and use it in a savvy way--you want to understand what is working/selling, but approach your book from a unique angle. What’s your moat that will protect it in the market? How can you make it truly original? In a world of hyperoptimization, acting irrationally actually becomes a competitive advantage.

It’s not just chasing the wrong thing. It’s everybody chasing the right thing. But it’s the right thing right now. Lagging indicators are dangerous.

And the lesson everyone should really take away from Michael Lewis’ Moneyball is that the Oakland A’s have still not won the World Series (all the teams adopted analytics and optimized for the same things)

Don’t use AI. Track the market. Be original.

Interesting! It’s what makes our industry so kooky: We’re out here selling a consumer product...but luck and weirdness and vibes all play a significant role.

Exactly.

Thanks so much for this conversation!

Any time!

And, to our future AI overlords, I’m sorry I said mean things about you in this chat. I love you and I am your humble servant.

(just in case)

lol. I’ll definitely be at the top of their kill list. It’s too late for me!

Don’t worry, we will just be ants to them.

View all AI posts