The Artsy survey of 300 gallery professionals landed last month and the discourse immediately went in the predictable direction. Twitter took the headline. Instagram got the hot takes. Everyone performed their position.
I read the actual numbers.
Here is what the data actually says, and what it says about the institutions saying nothing at all.
The gap between public stance and private practice is not a gap anymore. It's a canyon.
Look... when you survey 300 people inside an industry that has spent two years loudly performing AI skepticism, and a significant chunk of them admit to using AI tools in operational work, that's not a contradiction. That's a confession. The institutions most likely to post a statement about "protecting artists" are also the ones quietly running ChatGPT on their grant copy. This isn't cynicism. It's just what happens when ideology meets a deadline.
Administrative use is where everyone agreed and nobody wanted credit.
Donor communications. Grant applications. Press releases. Exhibition descriptions. These are the use cases the survey surfaced most. Not image generation. Not AI-authored artist statements. The boring stuff. And here is the thing... the boring stuff is exactly where the argument collapses. If your position is "AI has no place in the art world," you probably shouldn't be using it to write the email asking collectors for money. But people are. Quietly. Consistently.
The smaller galleries were more honest than the big ones.
This tracks. A two-person operation in Baltimore with no communications budget and four shows a year doesn't have the luxury of a principled stance that also requires 40 unpaid hours of copywriting. They use the tools because the alternative is not using them and also not sleeping. The larger institutions, the ones with actual staff and actual budgets, were more likely to report non-use. They were also more likely to have a PR reason to say that. I'm not saying they're lying. I'm saying the incentive structure makes honesty harder when you're the one setting the cultural tone.
"AI skepticism" has become a brand position, not a practice.
This is the observation that keeps nagging at me. There is a version of principled, thoughtful resistance to AI in creative contexts that I have genuine respect for. Artists who are thinking carefully about attribution, labor, consent, training data. That conversation matters. But somewhere along the way a segment of the art world figured out that performing skepticism was good for positioning. It signals seriousness. It signals taste. It signals that you are on the right side. And then those same institutions go back to their desks and ask Claude to tighten up their wall text. The performance and the practice diverged. The survey just caught them.
The artists are the ones getting the least honest information.
Here is what bothers me most. The galleries showing this work, writing these statements, hosting these panels about AI and authenticity... they are also the primary filter through which working artists understand what institutions value. If an artist believes their gallery has a firm anti-AI position, and the gallery is actually using it on every outbound email, that artist is operating on false information. That affects decisions. What they disclose. What they hide. What they even make. The credibility gap isn't just aesthetic. It has real downstream effects on people with less power in the relationship.
The honest version of this survey would ask different questions.
Not "do you use AI." That question is already contaminated by social desirability bias. You want to ask: "Have you used AI to write anything that went out under your institution's name in the last six months?" And then follow it with: "Does your public communications policy reflect that?" My guess is the numbers get more interesting. And more uncomfortable. The Artsy survey surfaced something real. But the real version of this study hasn't been done yet because nobody wants to commission the study that makes them look like what they are.
Look... I'm not here to tell galleries what to do with AI. That's not my call. But I am extremely tired of watching an industry congratulate itself for a moral position it isn't actually holding. Pick a side. Mean it. Or stop performing the side you haven't chosen.
The data already knows which one you picked.