the debate is over. galleries aren't arguing about whether AI-assisted work counts as art anymore. Artsy surveyed 300+ gallery professionals this year and the philosophical hand-wringing is basically gone. what replaced it is a much more uncomfortable question: how do you sell something when the entire market is flooded with undifferentiated versions of the exact same thing?
that's the real tension now. not legitimacy. economics.
a few observations from digging into this...
the volume problem is worse than most people admit. when a single model can generate 1,000 coherent images in an afternoon, the traditional scarcity logic that props up art markets collapses. galleries have been pricing work partly on the assumption that skill creates a ceiling on output. that ceiling is gone. one gallerist quoted in the survey called it "the jpeg flood." i think that's exactly right. you can feel it when you open any portfolio site right now. the visual language is already converging.
process documentation is becoming the actual product. the galleries seeing any traction with AI-assisted work are the ones selling the process, not the image. screen recordings. prompt logs. iteration archives. the finished piece starts looking almost secondary. look... this is strange, but it makes a kind of sense. if the output is reproducible, the thing that isn't reproducible is the specific human thinking that got you there. that's what collectors are being asked to buy. the lineage, not just the landing.
"AI-assisted" and "AI-generated" are being treated as completely different categories now. galleries are starting to price accordingly. work where a human made hundreds of judgment calls throughout... selective editing, physical manipulation, reprompting, collage into analog processes... is commanding meaningfully higher prices than work that's closer to prompt-and-export. this feels right to me. there's a real difference in cognitive labor between those two things and the market is slowly learning to see it.
the artists doing best are the ones with a pre-existing visual identity. here is the thing about AI tools: they amplify what's already there. if you had a distinctive sensibility before you touched Midjourney or Stable Diffusion, that sensibility can survive the process and you get to move faster. if you didn't have one, the tool gives you nothing to hold onto. you just end up producing beautiful, technically competent, completely forgettable images. the survey bears this out. the artists getting gallery placement are almost all people with documented practices that predate their AI adoption.
edition sizing is the pricing lever nobody is talking about enough. a few smarter galleries are treating AI-assisted prints the way limited-edition photography got treated in the 90s. strict edition caps. certificates tied to specific outputs. destruction of the source files, or at least the promise of it. it sounds a little theatrical but it's working. scarcity has to be manufactured somewhere if it doesn't emerge naturally from the process. this is just galleries doing what they've always done... constructing the conditions that make something feel irreplaceable.
the buyer psychology shift is real and kind of fascinating. collectors who were skeptical two years ago are less concerned about the AI involvement and more concerned about the artist's track record of taste. they're essentially betting on curatorial judgment. can this person find the 3 images worth keeping out of 10,000 generated ones? that's actually a skill. it's not the same skill as painting, but it's not nothing either. the galleries framing it that way seem to be having an easier time closing sales.
none of this is fully resolved. the pricing norms are still shaky and anyone telling you they've figured it out is either very early to a gallery relationship or not being honest about their sell-through rates.
what i do think is clear: the artists who are going to navigate this have a point of view that exists independently of whatever tool they're using. the tool is infrastructure. the perspective is the thing.
if you're building the business side of a creative practice around this kind of work... tracking editions, managing releases, handling the documentation side of things... LUNARI is worth looking at. it's built for solo creators who need operational infrastructure without a full team behind them.
but the art part? that's still entirely on you.