The art world has a position on AI. It is complicated, it is nuanced, and it gets delivered at panels in a very specific tone ... the one that implies the speaker has thought more carefully about this than you have. The concerns are real. The aesthetics of the conversation are impeccable. And behind the scenes, while those conversations are happening, the galleries are just ... using it.
Artsy's 2026 survey of over 300 galleries is one of the more quietly damning documents I have read this year. Not because it reveals some scandal. Because it reveals a gap so wide between what the art world says and what it does that you almost have to admire the discipline it takes to maintain the performance.
Here is the thing ... when you ask galleries what they think about AI, you get the thoughtful version. Ethics. Authorship. The sanctity of the hand. When you ask what tools they are actually running in their operations, you get a completely different answer. Email drafting. Wall text generation. Artist bio rewrites. Press release copy. Pricing research. Audience segmentation for fairs. The survey did not name specific tools, but the categories were specific enough. These are not experiments. These are workflows.
And look, I am not even criticizing this. The gap is interesting. The gap is the story.
What the survey captured is the difference between institutional identity and operational reality. Galleries have spent the last three years carefully positioning themselves relative to AI ... in public statements, in artist relationships, in how they talk to collectors who are themselves anxious about what AI means for the work they are buying. That positioning is real and it matters. But it has almost nothing to do with whether the associate director is using an AI tool to draft the Frieze application copy at 11pm on a Thursday.
This is not hypocrisy, exactly. It is something more structural. The art world has always maintained a fairly clean separation between what happens in the front of house and what happens in the back. The mystique is part of the product. Nobody tours the shipping crate room. Nobody sees the condition reports. Nobody reads the insurance valuations. The conversation about a work's meaning and the conversation about its logistics are kept rigorously apart, because one of them would contaminate the other if they touched.
AI just got sorted into the logistics pile. That is actually a coherent decision.
But here is where it gets complicated. The survey found that adoption was almost entirely concentrated in administrative and communication functions, which tracks. What it also found ... and this is the part worth sitting with ... is that the galleries doing the most internal AI adoption were not the ones showing the most AI-generated or AI-assisted work. The correlation ran the other direction. The galleries most vocally skeptical of AI as a medium were often the ones most quietly using it as an operational tool.
That inversion is philosophically interesting and also completely predictable. If your brand is built on the irreducibility of human making ... and for a lot of serious galleries, it is ... then the last thing you want is for your collectors to start wondering where the lines are. So you keep the tools in the back. You use them for the work that was always invisible anyway. You do not advertise it. And you continue to hold a nuanced position at panels.
I have talked to a handful of people who work at galleries in the mid-tier range ... not the blue-chips, not the emerging spaces, but the ones that have been operating for fifteen to twenty-five years and have a real program. Every single one of them confirmed some version of this. One person told me they use AI to generate first drafts of collector emails and then rewrite them entirely, because the rewriting is faster than starting from scratch. Another said their wall text process used to take three days per show and now takes four hours, with most of that time going into the editing. A third was using it for fair booth layout proposals before handing off to the director for final decisions.
None of them wanted to be named. Not because they are ashamed, but because the conversation is too loaded right now and they do not want to be the person who made it more complicated for the artists they represent.
That is a real and reasonable concern. Artists are watching how galleries behave. The trust is not abstract.
But here is what I keep coming back to. The 300 galleries in that Artsy survey are not outliers. They are the mainstream. And the mainstream has already made a quiet, practical decision about where AI belongs in their operation. It belongs in the invisible work. It belongs in the parts of the job that were always undervalued and underpaid and mostly done by overworked associates who had degrees in art history and were answering emails until midnight.
If AI is absorbing some of that load, that is not a threat to the art. That is a staffing conversation. And it is a conversation the industry has been avoiding because having it out loud would require admitting how much of what galleries do is just ... administrative. It would puncture the mystique in a different way than AI-generated painting does. Less dramatic. More honest.
The irony is that the art world is very good at talking about labor. It talks about artist pay, about fair compensation, about the economics of making. It is less good at talking about the labor conditions of the people who work inside galleries, many of whom are doing enormous amounts of writing and communication work for not much money. If AI is quietly reducing that burden, that is worth acknowledging instead of hiding.
Look ... the discourse will catch up eventually. It always does. Some gallery director will write a thoughtful essay about how they are using these tools and why, and it will get a certain kind of response, and then six months later it will be normal to talk about. That is how these things go.
But right now there is a 300-gallery survey sitting quietly in the Artsy research archive, and it says something the panel conversations have been carefully not saying. Which is that the decision has already been made. The question was never whether. It was always where, and how, and for what.
The art world sorted AI into the back room. That is a choice. It is probably the right one for now. But eventually the back room and the front room have to be the same building, and the people who work in both deserve a more honest conversation about what is actually happening inside it.