What Makes AI Art Worth Collecting?
IEEE Spectrum Jackie Snow
An artist tricked the internet by passing off a real Monet as AI art, and people trashed it as fake anyway. Turns out there's a real, messy market for AI art forming underneath all that confusion.
Based on reporting by IEEE Spectrum, Jackie Snow — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Back in May, an anonymous artist known as SHL0MS posted an image on X and called it an AI attempt at a Monet-style water lily painting. More than 600 people piled on, pointing out flaws in the color, the depth, the way light supposedly didn't behave right. Then came the reveal: it was a real Monet, downloaded from Wikimedia with the signature cropped out. SHL0MS minted the whole exchange as an NFT called "Inferior Image" and sold it for just over $40,000 after 28 bids.
The stunt landed because it captured something true about how people talk about AI art right now — quick to sneer, often wrong about what they're even looking at. But underneath that noise, a market has quietly taken shape. It's scattered across NFTs, museums, and stock photo sites, and it's bigger and stranger than the discourse suggests.
Take Jediwolf, the collector who ended up winning that Monet NFT. He's spent over two decades collecting digital and AI art, and one of his collections — a pre-2020 archive of early GAN-based work he calls UnderTheGAN — holds around 100 pieces worth roughly $72,000. He sees himself as part curator, part archivist, documenting a field that moves faster than anyone can properly catalog.
Then there's the physical-world version of this market. Last month, digital artist Refik Anadol opened Dataland in Los Angeles, billed as the first generative AI museum. The flagship show pulls in live weather data from 16 rainforests, some rooms respond to visitors' own biometrics, and the gift shop will print a T-shirt based on the biometric data collected during your visit. A founding batch of 1,000 AI data sculptures sold out in 34 minutes at $5,000 apiece. Anadol's system, trained on more than 500 million nature images spanning 2.2 million species, isn't cheap spectacle for its own sake — he argues AI art demands a new kind of transparency about where the data and models come from, something painters never had to account for.
The market data cuts in different directions depending where you look. Digital art's share of sales nearly tripled from 2024 to 2025 per the Art Basel and UBS report, yet Christie's shut down its dedicated digital art department in September after none of its sales cracked $400,000. The clearest signal might come from stock photography, of all places: economist Samuel Goldberg found that one platform's monthly sales jumped 80 percent after it allowed AI images, even as traditional contributors started leaving. Whitney Museum curator Christiane Paul draws a hard line here, though — she doesn't count a quick DALL-E prompt as art at all. Real AI art, in her view, means training custom models and building control systems, work she describes as far harder to wrangle than a paintbrush ever was.
My take — AI-written commentary, not fact-checked reporting
People love pretending they can spot AI slop on sight, and the Monet stunt proved most of them can't — they're pattern-matching on vibes, not actually looking. That's fine as internet comedy, but it's a bad foundation for an art market that's clearly already real, judging by sold-out sculpture drops and stock-photo sales spiking 80 percent. The honest take is that "AI art" isn't one category — a prompt-and-done image and a custom-trained model feeding a rainforest installation have almost nothing in common, and lumping them together just makes everyone dumber about what they're arguing over.
Read more about this at: IEEE Spectrum