If you think AI can endanger your art, I will come to your house and ask you to take yourself seriously

or: Works of passion in the age of mechanical reproduction

Jun 03, 2026

From a 1789 book attempting to explain the inner workings of a chess-playing automaton known as “the Turk”

It’s irritating to me that this even bears saying, but such are the times we live in: I do not use AI to write. Why would I? Writing is what I do in order to think. It’s immensely pleasurable, if more than a little compulsive. Because it’s the default response to whatever’s going on for me, it’s also a thing that can all of a sudden just happen: notebook out, laptop open, my wife mercifully spared yet another disquisition on why Zendaya can perform addiction like no other actor of her generation. I become like a gremlin if I don’t write. You would no sooner want to live with me in this state than you’d want to live with someone who half-prepares multiple fruit pies and then leaves them all on the counter to rot and the dishes to crust over in the sink.

Being like this means that I generate tons of words that are totally unusable. I also become exasperated with myself for neglecting longer, slower projects (my novel) for faster, dopamine-seeking ones (hi, Substack!) I struggle to this day to take the advice I’ve given students for years: Take yourself seriously. Even when writing finally pays the bills, that doesn’t mean you don’t deserve some down time. But none of these things — nor the various prostrations, humiliations, and checkings of my own ego that needed to happen in order to facilitate this life of gentlewoman’s graphomania — would be enough to make me outsource my writing to the robot. I mean, would a bibliomaniac be satisfied with such an outsourcing of labor? A kleptomaniac? (A weird example, but you get my point.)

Instead of using AI to write, I’ve had humorous and infuriating conversations with ChatGPT, whom I call Specula, because the thing really is a cracked lookingglass. For our purposes, I’ll just call it “the machine” or “the robot.” The machine and I have chatted about niche historical interests like Rose Kennedy’s adolescence and the gender politics of East Germany. I’ve solicited its opinion on hot-button topics and then argued with it. It responded to me first like a customer service agent, and then like a social worker: You’re right, Raf. I really overstepped my bounds by just considering the historical facts and missing how people might have been feeling during the Spanish Civil War. More than once – here’s the cantankerous junior prof coming out – I demanded it cite its sources. When this became too much, I asked it to draw surreal pictures for me. Visual jokes so incoherent that only I would find them funny, like this one about Mark Twain appearing on a manosphere podcast hosted by Spiderman:

AI generated image of Spiderman talking to Mark Twain on a his Inside the Spidosphere Podcast — My AI generated image of Spiderman talking to Mark Twain on his *Inside the Spidosphere* podcast

Three years ago, I was teaching at a public university with a colleague who was an early adopter of AI. This colleague had published several books of rural gothic realism, but he’d also been into game design in the ’80s and loved William Gibson’s Neuromancer. You can see where I’m going with this: he thought it might be interesting to play around with the iterative machine. Argue with it about Blade Runner versus Do Androids Dream of Electric Sheep? Ask it the meaning of life and see if it would cough out “42.”

I landed that job because I managed to express fluency with what we’d decided to call “digital narrative,”1 I resisted all the VR helmets and AI chatbots my colleague wanted me to try. I decided I’d be using a little thing called postmodern theory to address the vagaries of cyberspace, thank you very much! Who needs to put on a helmet and visualize a vast pixelated desert when you’ve got a stack of books by chainsmoking Europeans about how that pixel-desert represents both hyperreality and the death of the real?

Yes, it was a lot of fun to analyze vaporwave videos with students and hyperbolize about whether we’d arrived at the end of history, but all this philosophy was just another model of reality, no more real than the cyberspace the philosophers were wringing their hands over. And when I took a look at how my colleague was using AI in his creative writing classes, I learned a thing or two.

His students were using LLMs to build fictional characters from their novels and chatting with them. They were prompting image generators to create a scene or setting from the story they were writing as a way of spitballing ideas. When they liked the dialogue or image the machine produced, they tinkered with it, seeing how it might inform what they were writing. When they didn’t like what they saw, it was an opportunity for a refinement of what they were imagining.

If the machine produced something uncanny, unpleasant, or otherwise representationally dubious, they researched why. Whence the robot’s racial and gender biases? Why are its metaphors so bizarre? Why might it struggle to “understand” certain character motivations and not others? The robot may have been a collaborator, but the students were the brilliant draftspeople. There were no illusions about who was authoring their stories; likewise, there were no illusions about the engineers who’d authored the machines.

When you study the moves the Mechanical Turk makes, you end up learning a lot about the dude hiding under the table. The same is true for any artwork and its artist. It’s an interesting, even enlightening exercise. Another word for it is “criticism.”

I was impressed with my colleague’s approach, and with the student’s results. I caved and got a cheap subscription to an AI image generator. It was fun to feed it prompts and see what it could conjure. “Girlboss wellness influencer going live while crossing River Styx,” or “Schlubby millennial in Shakespeare & Co. showing 1920s Left Bank intellectuals hilarious video on her phone” or “mermaid cat on pink cloud thinking about God and the universe.” This was not artwork I identified as my own, not artwork I had any illusions about “creating.” I was no sorceress, just a witch stirring a pot and observing the vapors.

The robot may have been a collaborator, but the students were the draftspeople. There were no illusions about who was authoring their stories. There were no illusions about the engineers who’d authored the machines.

I started this Substack when I decided to leave academia. I was, for reasons that will be familiar to anyone who’s emerged from addiction, teetering on the brink of financial bankruptcy. I had pie in the sky hopes that the Substack might improve my situation.

Sometimes, I’d let the robot give me writing prompts. It would work like this: A sentence would pop into my mind (e.g. Reports of my cancellation are greatly exaggerated). If it made me chuckle enough, I’d prompt the robot to draw a picture of it (e.g. “Mark Twain reading a blog while angry mob forms behind him”). Sometimes, just watching my through-form turned into an image was enough. Others, the image would become a prompt: I’d write an essay about how we’re so afraid of appearing contrarian that we won’t even allow ourselves to take in opposing viewpoints, let alone try to understand them. I asked the robot to draw me “an invincible heart,” and then I wrote this essay about the 2024 election; I asked it to draw me “the anti-fascist protestor is also the hyper-militarized cop,” and then I wrote this essay about moral responsibility; I asked it to draw me “the sad man from the cover of A Little Life as a beat-up hockey player” and I wrote this essay about fujoshi fiction.

It was a sense-to-nonsense-to-sense pattern that I’d never tried with my writing before, a Dadaist way of organizing my thoughts through deliberately disorganizing them. The robot was so-so as a research tool and middling as a conversation partner but kind of amazing as a funhouse mirror of my subconscious, if not our collective subconscious. Audrey Horne put this elegantly in a recent essay on AI:

Languages are the models or maps we’ve built to describe the territory of reality, which is made up of the material and the immaterial. The material and the immaterial have a certain shape to them, a shape we all bump into enough that we can agree to talk about them in a certain, shared way. We stumble into the same shapes and relationships, across time and across cultures.

And here’s Horne on how the way the robot mirrors our concepts might make this more apparent to us:

Physicists dove into matter and found — at the subatomic level — that it was made of something ungraspable, something that would only solidify when it was observed. Maybe LLMs are revealing something similar in reverse: that the immaterial, viewed from far, far up, has some tangible quality. When pushed, matter dissolves and the abstract solidifies. Matter and the abstract are moving toward each other.

The philosophical implications of this are incredible. But unfortunately, we’re still trying to sort out the pragmatic implications of the machines among us. There’s the proliferation of online slop and offline scandals like the slop-novel Shy Girl and the AI-generated winners of the Commonwealth Short Story Prize. There are the accusations of mechanization among writers and readers alike, the typical furor and sanctimony, not to mention the cursed LLM syntax like the most insipid 90s ad copy. Who among us hasn’t felt that nervous catch in the throat while reading an essay you’ve been enjoying only to stumble upon a couple clauses like this: If anything, you’d probably be feeling like your skin is an inside thing accidentally exposed to weather. And that matters. What?!?

It was hard not to constantly hear the voice of “the demon,” as critic Sam Kriss describes AI in a recent essay about the omnipresence of the LLM sentence. It was hard not to feel haunted by the no-there-there of automated writing, hard not to think about how many people were outsourcing their thinking to the machine, passing off its zero-qualia hallucinations as the product of deep thought. I read more about the massive data centers and the metric tons of water.2 I started getting eerie AI-generated scam emails from famous authors who somehow knew who I was and loved my books. I switched to writing in Scrivener so I wouldn’t be badgered by Word’s obnoxious new AI tools (No, I never want your help “professionalizing the paragraph”!!) and I was part of the Anthropic class action suit. It was all maddening and deeply unsettling.

“The distrust and anger right now are just way too intense,” a young artist told me as we discussed their powerful dislike of AI. “It’s just best to quit putting out pieces with any gen AI.”

I understand exactly where they were coming from: it’s a matter of both principle and pragmatism.3 Financial solvency finally permits me to run this Substack like a publication, on a schedule, for which I deliberately commission original art – or even make my own in a pinch.4 Given the abundance of both slop and slop-paranoia, I want my readership to know, beyond a shadow of a doubt, that it’s always me you’re dealing with, not some vacant machine blinking behind a rubberized Raf mask. Nothing deflates the reading experience and sows distrust quite like recognizing the machine’s babblings being passed off as human qualia. Nothing degrades art so readily as the assumption that an art-like thing can be mechanically reproduced and still function in the world like a work of art.

Still, anxiety over slop has reached such a pitch that even having posted my little machine-image-prompted essays with images included now feels like an uncalculated risk. I can see in the AI discourse shades of the moral panic of the 2010s. That long, impulse-driven Dark Age when we were all freaked out all the time, when anyone among us could have been sniffed out as a racist or transphobe or class traitor and pilloried for our trespass. Back then, when a problematic fave was up for cancellation, the question was almost never, “Has the artist produced an artwork?” It was, “Has the artist produced an artwork while being morally unimpeachable?” More specifically, it was: “Does the morally dubious artist deserve all this material success?” If the question was being asked, the answer was always no.

In his anti-AI polemic – which doubles as a death threat to all who would pass off AI as their own work – Kriss correctly identifies the AI output as meaningless: “what they speak is the language of angels, in which, like the chirping of birds, there is neither truth nor lies.” Yes, you cannot impute semantic intention to an AI sentence: there is no consciousness to answer for the garbled metaphors, no physical body to explain an LLM’s assertion that X feels like Y. Yes, it’s annoying that the language is everywhere (including, as Kriss points out, in the pilot episode of a limited series called The Miniature Wife). But – and this is crucial – the robot is not producing art. The robot is not becoming passionately obsessed with something and needing to get it down on paper. The robot is not seeing a Krispy Krunchy Chicken sign with the first K blinked out and imagining an entire scene unfolding in that gas station parking lot, totally unbidden. The robot is not reading the Elizabeth Bishop poem “The Fish” and remembering how its first boyfriend used to squint at it through a toy telescope.

So how does the robot threaten the artist? In terms of clout, it would seem. Here’s Kriss on the screenwriters of The Miniature Wife:

The writers and showrunners of The Littlest Wife (sic) are called Jennifer Ames and Steve Turner. They are apparently quite a powerful pair; they previously worked together on Boardwalk Empire. I don’t know exactly how much they got paid for ‘nursing the burn of every sip,’ but I’m sure it was plenty. Enough to feed an entire village of semi-literate Cambodian peasants for a hundred years. I’m just not sure why they should get that money, when their job could just as well be done by any one of those semi-literate Cambodian villagers. This is the democratising power of AI I’ve heard so much about, isn’t it? If you’re a pair of lazy, talentless hacks, gobbling up opportunities more interesting people would kill for and then not even bothering to write the show yourself, sooner or later, the people who sign your cheques will cut out the middleman, and you will deserve it.

Kriss isn’t wrong about the dishonesty or the mechanization here, but I can’t help thinking that he’s talking about an issue in the art and media worlds that predates AI. Hasn’t there been pabulum written by committee being broadcast on TV for literal decades, Golden Age notwithstanding? Is passing off the robot’s writing as your own and getting paid for it the same or different from passing off another human’s? Long before Shy Girl became the poster child of novelistic dishonesty, there were phoned-in novels by massively successful writers, towering scholarly works “typed” by wives, and professors who sampled liberally from the work of their students. And if we’re talking about cobbling together a literary persona out of lies, we could look past the 2026 Commonwealth Prize and into a not-so-distant past that includes Stephen Glass, James Frey, and JT LeRoy. Or maybe we don’t need to go that far back: what about the bestselling 2025 memoir The Tell, in which billionaire Amy Griffin appears to have used not AI but the pseudoscience of recovered memories and another woman’s trauma to tell a harrowing story of childhood sexual abuse?

The issue is not that the robot is producing the art we humans could be producing – and even if it could, our collective imagination doesn’t work like a zero sum game. The issue appears to be that humans now have a more efficient route to becoming rich, praised, and famous by outsourcing the production of schlock. Theorist Tony Chamas (Tony of 1Dime) puts it simply: “AI is only a ‘threat’ to real art because the standards of what is considered ‘art’ have already been so thoroughly lowered due to the process of ‘standardization,’ whereby art becomes standardized into mass-producible cultural commodities designed for maximum profitability.”

Art cannot be replaced with the babblings of the machine, but “commercially successful” art – i.e. that which has been pre-engineered to appeal to a broad-as-possible consumer base, otherwise known as “entertainment” – can be imitated by it. Whether writers will be replaced isn’t what’s at stake here. What we actually appear to be panicked over is how the greedy oppressive oligarchs are getting theirs while us truth-and-beauty-loving victims are getting so little of ours. Which means that again, we will start crusading. We will wish for the downfall of the “lazy, talentless hacks,” we will issue death threats in hopes of rooting out the covert oppressors hiding among us.

I know humans cut corners and plunder resources and do bad things that impact other humans at scale, but I can no longer summon that sweaty, righteous secret police energy that had us all surveilling each other for signs of Secret MAGA Nazism circa 2018. I want to continue taking the trust of my readership seriously and commissioning artworks from human artists: these are pragmatic decisions, and pro-social ones. What I don’t want to do is compulsively feed text into Pangram, await the Gotcha! with bated breath, and then Schadenfreude over yet another public figure’s downfall. I feel so richly done with that shit.

The thing about moral panics is they really tend to miss the point. While we were busy sorting out which of our rich and famous faves were most problematic in 2013, we were notably not writing the novels or painting the canvases that might have made us someone else’s faves. While we were busy fretting over book sales, trying to game our way off the midlist, and snarking about romantasy in 2021, we were not taking to the page to indulge our riskiest and most exciting ideas. We were letting the thought-forms evaporate. We were letting our minds spin.

Inspiration is integral to art-making. In some cases, the robot can give it, but there is no case in which robot can take it away. It’s the writer’s remit to get her inspiration down into notebooks and onto index cards. It’s her job to carry it out into story form. This is no easy feat. It requires clearing the following hoops: 1) You’ve got to take the idea seriously; 2) You’ve got to take yourself seriously; 3) You can’t feel so resource-precarious that writing feels irrelevant; 4) You can’t feel so irrelevant that writing feels resource-precarious.

This is why I’m as unbothered by a writer who takes inspiration from the robot as I am by Marcel Duchamp’s Dadaist Fountain or a straight writer who becomes interested in writing a gay character. Acting on inspiration takes work. Inspiration is different from theft. Inspiration is different from deception. Inspiration can come from unusual places. Real art cannot be faked, and the best art is adventurous, risky, a little culturally contrary.

There is certainly such a thing as deception. There is also such a thing as phoning it in for clout. But there is no such thing as “purity of inspiration.” If we re-enter that dangerous loop where we demand moral perfection of an artist or artwork, we will once again create an environment hostile to the production of artworks altogether. Crusading against the robot and its beneficiaries may feel thrilling for a while, but expose enough grifters and eventually years will have gone by and you still won’t have written your own novel. What you will have done is stoke your paranoia.

1) You’ve got to take the idea seriously; 2) You’ve got to take yourself seriously; 3) You can’t feel so resource-precarious that writing feels irrelevant; 4) You can’t feel so irrelevant that writing feels resource-precarious.

What about this instead? Let’s all take ourselves more seriously. Let’s do a thought experiment where we pretend that instead of being “bitter and frustrated gremlins…terrified of our own mortality” as Kriss puts it, or unsuccessful/underselling/unloved artists grasping at relevance, or a bunch of sad little Davids battling a horrible greedy cabal5 of Goliaths, we are simply people moved to put words to paper because we enjoy it, are prima facie worthy of doing it, and trust that if we keep doing it long enough, other people will want to read what we’ve written.

The robot’s presence is currently annoying and sometimes ethically dubious. The robot’s sentences are maddening. But the robot cannot write the novel that you could be writing right now. Crazy as this may sound, the robot also can’t prevent you from publishing a book, or building an audience. It can’t execute judgments like “I am bad and wrong. I shouldn’t be writing.” If that thought’s in your head, it originated with you. And now I – a living, breathing human being typing this sentence – am asking you why you’re thinking that.

Here’s a real human writing prompt.

If you want to meet other real human writers who hate fishbowl workshops and love inspiration, here are my classes in fiction and nonfiction.