That’s the Deal: What Rankles You Most About AI Prose?

How the Gift Economy Explains Our Disgust with LLM Writing

It’s a good thing the brouhaha about AI writing is finally blowing over. I keed, I keed; it’s hotter than ever. “With a new AI scandal engulfing publishing seemingly every month,” writes Anna Silman in the Wall Street Journal, “it’s become more difficult to punt questions about its impact into some distant future.” No foolin’.

Writers, agents, editors, publishers, retailers—the entire industry is pretzeling itself trying to figure out what to do and how to think about the issue. And that’s just book publishing; as you all know, the same dynamic is playing out here on Substack, LinkedIn, newspapers, and other venues.

Photo by Joshua Rawson-Harris on Unsplash

Jumping headlong into the mess, Giles Crouch, PhD-c offers up a fascinating take involving the social contract between authors and readers, focusing on anthropologist Marcel Mauss’s concept of gift economy. “Even if we can’t always detect AI writing,” says Crouch, “when we learn that it is, we are disappointed. Why? It’s the gift economy”—about which more in a moment.

Questions of Style

AI slop may be lousy for aesthetic reasons, though I’ve read plenty of work I know for a fact was written by AI that was perfectly serviceable for what it was. You probably have too, even if you couldn’t tell. In fact, beyond obvious tics and tells, AI writing can be somewhat challenging to spot, as various studies and stunts have demonstrated.

“AI prose is distinguishable by word and punctuation choice as well as sentence and paragraph structure,” writes the Economist, reporting on the findings of their own internal study. “But its hallmarks are not what you might expect, partly because its writing style has changed with software updates.”

The overuse of em-dashes, for instance—Claude still overuses them compared to humans, but ChatGPT seems to have hit peak em-dash use in 2025 and has since throttled back its use. Weirdly, LLMs tend to underuse punctuation, according to the Economist’s study. “They use fewer commas and semicolons than humans (and hardly any parentheses).”

The supposed hallmarks of LLM prose aren’t terribly hallmarky if they change, rendering detection harder. Sometimes we can spot AI prose, sometimes not; that’s why Substack has partnered with the detection outfit Pangram. But even AI writing that is somehow excellent aesthetically—or at least not quirky in off-putting, head-scratching ways—still leaves us feeling meh or bleh when we discover its provenance. Why?

Crouch points us to an answer.

The Deal on the Table

“The idea of the gift economy is fundamental in how we create social bonds, form and maintain relationships, deep or just transactional, such as with a brand through marketing activities like advertising,” says Crouch. “Writing poems, sonnets, novels, non-fiction books, articles, blog posts, all are forms of this moral and social gift-giving. The writer creates a book and publishes it because they want to tell a story. When we read the book, we are reciprocating the gift by reading it.”

A piece of writing is a pact between the writer and reader. It’s a contract, and we feel betrayed or cheated when we discover the writer didn’t invest their work with the effort, intent, or skill we expected. It’s like buying a bag of chips and finding out the package is mostly air. We feel ripped off.

Deepening this dynamic, Ruth Gaskovski and Peco recently discussed the “I” of the creator. “There is an ‘I’ behind a work of art,” they write. Ruth mentions her first encounter in the Louvre with Monet’s painting The Luncheon, examining the brushstrokes. “Even though I was well over a century removed, I felt that his hand was somehow within my reach, hovering in the same physical space that I was now occupying.” She was communing with the artist’s residual presence in the work.

But what if there’s no real artist involved? What if the giver is passing off something other than the gift we assume we’re getting? Our answers to those questions are behind why a book gets canceled or a deal falls through when AI use is flagged or suspected; it’s not just about cringy writing or copyright issues (AI-generated prose can’t be copyrighted, so it buggers the publishing model). It’s about the feeling that there’s no “I” on the other side of the exchange. That’s part of the let-down when we find an artist has used AI in the production of a work.

The Aesthetics Issue

In this artificial moment, humans crave authenticity, signaled in some cases by the errors humans are prone to make. “When I see something that’s honestly badly written, it’s such a breath of fresh air now,” said one literary agent interviewed by the Journal. I’ve encountered—and maybe you have too, maybe you think this yourself—countless people on social media saying they cherish the typos in their or others’ work because it reflects the humanity of the piece.

I can’t get with this notion. I find it a little baffling, in fact.

When it comes to my version of the deal on the table, the unspoken pact between a writer and me, I mostly care about the final product and whether it’s any good. I’ve made this clear in all sorts of ways at this point, but my primary concern is the aesthetic experience of the work.

If I found out an author I loved used AI to create something I loved, I wouldn’t feel betrayed; I’d feel impressed and intrigued about how he used the tools to arrive at something worth my appreciation. If he lied and said he didn’t use AI, then I’d be annoyed and disappointed—but I’d probably still appreciate the work.

In this regard, it’s similar to why I can enjoy a piece of work by a scumbag. I’m predisposed to separate the art from the artist. I still laugh at Louis C.K. when he’s funny. I also refrain when he’s not. I’m judging the artifact, not the artist. It’s the same with writing—of any sort. I don’t care how sincere or well-meaning a writer is if I regard the work as lousy.

When I worked at Thomas Nelson, we—editors, publishers, marketers, salespeople, publicists, executive assistants, everyone—were always boosting our authors and their books inside the building. As we moved about the day in and out of meetings, at lunch, whatever, conversations often involved some measure of raving about what we loved that the company was publishing.

Sometimes the buzz about a certain author might be impossible to avoid. Many of my colleagues urged me to read one of their then-current favorites, a bestselling suspense novelist. I did. Even trying to be charitable, I could barely stomach the stuff. The stories were simply horrible—“to,” as my daughter would say, “my tastebuds.” All that good will from my friends and however much “care” and “craft” the writer employed could never overcome the artistic demerits of his work, which were legion. I tried; I couldn’t.

Something in Return

Now drag my revulsion forward twenty years. Would it have disappointed me to discover this novelist had used AI to write? No, though it might have enhanced my revulsion in one regard: Better tools should have improved the work; if it didn’t, my annoyance at having wasted my time would have likely increased. My sense of betrayal was grounded in the hope I would read something good that wasn’t.

I share my stance to acknowledge something here, something clear from Crouch. There is a contract in place when it comes to reading. If I’m going to give my time and attention, I want something in return. We all do. Reading is a trade. What that something is might vary—every reader is different—but we’re not trading our time for nothing.

If, for instance, I’m reading a memoir, I’d feel betrayed if I found out the author lied about their life; it wasn’t their life. If I read a mystery, I’d feel betrayed if the solution were obvious, an aesthetic failure. Those two examples differ in that the first involves the honesty of the author and the second their skill; there could be other forms of betrayal as well, as many as there are sets of expectations.

The challenge is that the contract between writer and reader isn’t explicit, and we all approach a work with our own hopes and assumptions. Some people, for instance, feel betrayed when they find out a ghostwriter was involved in a book; ask anyone in the industry about it, and they’ll roll their eyes. Of course, a ghostwriter was involved, they’ll say; the author can’t write for crap. But not all readers know hiring ghosts and rewriters is standard practice and so feel let down when they realize an author didn’t “write” their book.

AI writing is now forcing us to clarify the contract in other ways. “Publishing has always been an industry rooted in trust,” writes Silman. “Editors rely on authors to attest that their works are both accurate and original.” It’s a big hassle when a work is not. I’ve had to pull books for bad scholarship and plagiarism both.

But it’s not just editors and publishers. The reader experience relies on implicit trust as well. That’s why James Frey got into trouble. It wasn’t just his publisher that got flustered over his fabrications; readers did too. You’re telling me this was fake?

Clarifying the Contract

With the advent of LLM writing, a significant number of the reading population wants assurances what they’re reading was written wholly by a human. It won’t be exactly easy to sort this out since writing is a complicated process that can involve all sorts of inputs originating outside the brain of the writer. What’s more, it’s always been technologically assisted—writing itself is a kind of technology that has incorporated other technologies as they’ve been developed.

C.S. Lewis disliked typewriters. Gore Vidal hated word processors. As we’ve made peace with those developments, will we also acclimate to AI prose? Economist Tyler Cowen says yes, and he’s probably right to one degree or another. Expectations change, which means the terms of the contract will change.

Which matters more, authenticity or aesthetics? Or will we arrive at some new negotiation between the two? Surely, the answer isn’t we solve for one at the expense of the other. Give me typos and bad writing and I’ll toss your book. I also imagine there will be considerable diversity in how we ultimately navigate it all. Some readers will insist on 100 percent human, verified by whatever means are available. Others won’t care. God only knows how this develops.

The value of Crouch’s insight is that the gift economy will work regardless of how we decide. As readers, we’ll develop our own standards (various as they may be) for what we consider acceptable and not, and the writers we read will have to work to those standards if they want to be read.

That’s the deal.

For what it’s worth, I used Claude to proofread this post before posting. It caught eleven typos. Update: After I published, I caught one I introduced while fixing another. Gah!

---

Thank you for reading! If you enjoyed this post, please hit the ❤️ below, 💬 discuss, and share it with your friends—or anyone you know who prizes human art.

Not a subscriber? Take a moment and sign up. Free subscribers get two-week access to all new posts. Paid subscribers get access to the full archive, going back to 2022, plus an exclusive member chat and monthly member calls.

Before you go, check this out👇

The Wrong Fight over AI and Writing

Bookish Diversions: Use AI, Lose Your Book Deal—and Maybe More