Brandon Sanderson’s Case Against AI Art

I’ve spent a significant chunk of my life reading—and, more recently, listening to—fantasy and sci-fi. That’s probably why I clicked so quickly when YouTube recommended a video of Brandon Sanderson, an author I admire (and who, if I’m honest, is too productive for me to keep up with), talking through his take on AI art. The video is very good, and I highly recommend you watch Brandon read his essay, even though it contains no chasmfiends or mistwraiths.

Brandon’s essay shows intellectual humility, which I find lacking in the AI Art discourse (a genre that tends to oscillate between victory laps and moral panic). He starts talking about how Roger Ebert has infamously said that “video games are not art.” A stance he and his nerdy audience take as prima facie false. He goes on to note that Ebert is no pioneer in calling bullshit on whatever new form of art emerges: earlier iterations included poets dunking on prose writers and portrait artists dunking on photographers. Thus, he wonders: Is he just dunking on the new thing?

He ultimately concludes that this is not the case. And the way he gets there is by clearing away the objections that aren’t doing the real work for him. Yes, he’s worried about the economics. Yes, he’s bothered by the ethics of training on artists’ work. Yes, the environmental costs matter. But he insists that even in a world where models are trained only on consented data and run on efficient hardware, something would still feel off.

The core of his argument is that AI art collapses art into product. Where he argues that art is also (and maybe primarily?) process: the activity through which a person becomes capable of making the thing. He illustrates this with his own early, not-great juvenilia: books that, for him, worked as “receipts” of him becoming a writer. Art, thus, is the process happening to the person making it.

He ends by concluding that this is why his take on AI art is different from Ebert’s take on video games. Because AI art is fundamentally different from the previous revolutions. And, on a more positive note, since art is what society collectively decides to treat as art, there’s a simple way of “winning” the war against AI art: it suffices that some people decide this form of art is not worth pursuing.

I’m sympathetic to the spirit of this—and still not convinced by the conclusion. I see three problems.

First, although art is socially constructed, that doesn’t mean it’s easily steered. I really wish Brazil still produced music like in the 60s or 70s, and I’m sure a lot of people do. But our wills are not enough, and that stuff only rarely fuels a banger TikTok. Cultural tastes move due to people’s will, but also due to what is cheap, frictionless, and easy. So even if we grant Sanderson the premise that art is what we collectively treat as art, it doesn’t follow that “collective deciding” is a crisp process we can just execute.

Second, I’m not sure I fully buy the historical analogy. Every revolution looks uniquely alien. When photography emerged, one could claim that photography removed the process, just like AI did. And if photography didn’t kill painting, it definitely reshaped it. Painting survived by shifting its aims: toward impression, abstraction, expression, and the kinds of seeing that a camera didn’t naturally deliver.

Third, Sanderson implicitly argues that there is no process in writing with AI art. But isn’t writing with AI akin to painting with a camera? If you define “process” as the physical act of applying pigment, then yes, photography looks like cheating. But if you define process as the chain of choices that expresses taste and intention, photography is saturated with it. So why, in principle, can’t AI art be the same? I understand that the slop being sold on Kindle feels subpar, but so did the first photographs when compared to the paintings of their time.

None of this refutes Sanderson’s insight: art is more than the artifact—or as he’d put it, “journey before destination.” But it does complicate his account of why AI art “feels wrong.” Maybe the bitter lesson here is that, when the time comes, the new and the strange will always find ways to disappoint us. Still, he need not worry about me: 2026 is the year I finally finish Wind and Truth.