On AI, in EPOCH and Elsewhere

This piece from our editor J. Robert Lennon is running concurrently on LITERAMBIVALENCE, his blog.

For some time now, I’ve wondered if I ought to join the ongoing discussion about the use of generative “AI” in creative writing. I have strong feelings on the subject, I teach writing, and I edit a literary magazine; at the very least, I reasoned, I ought to be on record articulating what I think.

But every time the impulse strikes me, I poke my head up over the parapet to see what others have written, and typically there’s already a chorus expressing exactly what I think, more deftly than I would or could, and I decide that the issue is already being adequately addressed.

Another thing has held me back: I feel that what I want to say should go, in a perfect world, without saying. That is, no real writer needs to hear it from me, because they already know it, and if they don’t, they aren’t a writer. By now you probably understand that I think there is no legitimate use of generative AI in creative writing, not for research, or outlining, or “generating ideas,” or of course composing. If you ask an AI to do any of these things, then you didn’t do them, which means that you didn’t write, and so what you did is definitionally not writing, and you are not a writer. In short, no writer can ever use AI, because the use of AI disqualifies one from the title. I assumed, falsely I guess, that this opinion was universally shared, and that no one I regarded as a writer would even consider using AI, because all of us like researching, outlining, generating ideas, and composing, and would rather die than endure the humiliation of having a machine do it for us.

(I also shouldn’t need to list the reasons that I think the technology is bad, but: it is trained on stolen material, it is despoiling the environment, it is making billionaires richer still, it has caused the price of digital storage and RAM to vastly increase, it has doubled energy prices in every commuity that hosts its data centers, it cannot innovate, and it is rendering whole swathes of everyday society and culture generic and boring, from your neighborhood bar’s new logo to the obituary column of your local newspaper to the vows of your least beloved cousin’s wedding. I do understand that LLM technology has been transformative in some sciences and in computer programming, and may ultimately be worth its cost to those fields. But for ordinary tasks that require self-expression, it is bad; and for art, it is simply anathema. So, like I said, as an actual artist, I figured I could ignore it. As a professor I can’t, of course, but we do have techniques for returning our students to basic principles, however labor-intensive these might be. As a writer and editor, though, I assumed myself to be in the clear—and should anyone submit AI-generated text to my magazine, well, I ought to be able to tell, right? It would be too generic and boring to pass muster.)

Then, a couple of months ago—and this is what has driven me to write about AI—a friend notified me that a piece of writing I had already published in a recent issue of EPOCH had been plagiarized from one of theirs, and they further believed that the plagiarist had simply plugged their work into an AI and asked it to change certain details. My friend even provided a list of comparison points between the two, as proof.

I mournfully agreed that the piece had been plagiarized. But at the time I didn’t think it could possibly have been assisted by AI—or, rather, I didn’t think that AI could do that. A few months on, though, I now know that AI can do that, because I’ve seen it done multiple times. You probably have too. AI can generate a poem, an essay, or a short story that fools the experts, and there’s no going back.

What to do? As editor, I’d love to be able to promise you that no AI-generated text will ever again appear in EPOCH. But it will probably will, because there is literally no way for us to know whether something is or isn’t AI-assisted, or even AI-generated. We could test our submissions with AI detection software, but it’s famously unreliable, and is, in fact, itself AI and full of shit in many of the same ways as the text generators. We could try to dedicate ourselves only to the strange, the original—to poetry and prose unlike any we’ve ever seen! But of course novelty is not the same thing as excellence, and I am not confident an AI prompt can’t result in something peculiar enough to fool us. (Its present default mode sounds like Ocean Vuong having a stroke; who knows what else it might be capable of.) We could read—and retain, with perfect recall—every half-decent piece of literature ever created, and instantly recognize which one an AI submission has plundered...but, sorry, no. (I didn’t even consciously recognize my friend’s work in the plagiarized doppelgänger, despite having read it before; if anything, subconscious echoes of the original may have actually helped me to appreciate the fake as something that “sounded right,” strengthening its veracity in my mind.) We could publish only those things that seem the most honest, the most true, that move us the most, with the expectation that algorithmically generated text could never trigger real feeling in us. But of course it can. I was deeply fascinated—and disturbed—by the first AI-generated video I ever saw, a Google DeepDream abstraction, in 2015. I was even moved by that repellent “You Know the 80s Miss You” nostalgia video that dropped in the summer of 2025—the one that first introduced people to the idea that persuasive AI video was possible—even as I recognized what an offense against filmmaking it was. I wept, you might say, through my retching.

Why? How? Human beings are made to be moved. We respond to patterns. AI generates patterns. When there is nothing of substance there, we use our imagination to fill in whatever will give us what we need. It’s no crime to be moved by it, no more than it was a crime for that orphaned monkey to clutch Harry Harlow’s wire mother.

The real crime, of course, is to approach the creation of art, and thus its recipients, with such contempt that you’re willing, even eager, to elicit emotions in people through no effort of your own. The screenwriter Lila Byock put it beautifully in a recent post on Bluesky: “There’s an implicit contract between artists & audience that the artist will expend significantly more time & effort to create their art than the audience will expend to consume it. AI art flips that equation, which is part of why it feels inherently offensive.” The message of AI-generated “art” is: You’re so stupid that I can make you cry without even trying. This is why the fascists love AI, and use it for all their propaganda: it is fueled by the bad faith they adore, it devalues the experts and expertise they hate, and it strengthens their conviction that people are rubes who deserve to be deceived.

And I think that, ultimately, this is why we are seeing a sudden rash of AI players in the literary landscape: its practitioners despise the “literary world” as presented to them in grant and award announcements, in hyperventilating paid social media promotions, in those syntactically convoluted Publisher’s Marketplace book deal announcements—that is, as a contest for attention. AI “writers” note the familiar couple dozen people who make five figures delivering one-hour readings at universities, and recognize the arbitariness of prestige, and figure they might as well get some of it for themselves. And if the ruse works, they’ve not only shouldered their way past the gates, they’ve proven that the entire endeavor is a scam.

If it seems that I rendered that previous graf with a suspicious level of joie de vivre, yes, you can be sure that I also dislike the existing system of literary promotion and prestige, sometimes to a psychotic degree. It has nothing to do with art, which is what I tell the occasional student who is so disillusioned by the machinations of litbiz that they have lost the desire to write at all. Reject it, I tell them. Write for yourself alone. Print up zines. Make stickers. Start a press. In lieu of grants, apply for a shitty job you can steal time from for your real work. Fuck, in other words, everything but art.

Proponents of AI, of course, take the opposite position: fuck art, then cheat to beat Big Lit. They can do that—I’m sad to say that no one is going to stop them—but their efforts only serve to expose the shallowness and cynicism of the prestige economy, of which they are now active participants, perhaps even the most worshipful of its penitents. You could call that a useful outcome, if it weren’t, in the process, hurting so many people who earnestly and productively engage with and benefit from that economy, myself included—yes, I teach in an MFA program and edit an Ivy-League-bankrolled litmag, despite the flaws of the system that makes these things possible. That system is, if nothing else, predicated on the idea that art is worth making and can be trusted as a true expression of some real self, values I think are worth teaching, despite my reservations.

But AI has broken that trust for good. It is presently worming its way into everything (a friend, proofreading this essay for me, said that the AI in her PDF reader interrupted four times with suggestions), and its products can never be disentagled from the genuine. I will never use it, of course, because I like making things, and I don’t hate myself, even when the things I make don’t turn out so good. If you hate yourself and everything you do, you will love AI, because it negates the self, and when you use it, you are doing nothing.

But please keep your nothing to yourself. EPOCH doesn’t want it. Clutch your own wire mother, and keep your self-disgust away from my magazine and my students.

— J. Robert Lennon, editor