Stanford's Graduate School of Business has run the numbers on what happens when AI-generated art shows up in a marketplace that used to belong to humans, and the numbers say what everyone already suspected but wanted a professor to confirm anyway: consumers do great. Everybody else, less so.
The Study, and the People Who Ran It
The paper is called "Generative AI in Equilibrium: Evidence from a Creative Goods Marketplace," authored by Samuel Goldberg, an assistant professor of marketing at Stanford GSB, and H. Tai Lam of UCLA. They looked at a large stock-image and video marketplace — nearly 500 million images and videos on offer, a sample of 3.2 million of them tied to 62,000 individual artists and producers — and tracked what happened after the platform let AI-generated images in, starting in December 2022.
This is, dear reader, an actual controlled before-and-after comparison, not a vibe. Economists ran it. Numbers came out the other end.
The Numbers
Once AI images were allowed: monthly image volume rose 78 percent. The count of active firms — sellers, essentially — jumped 88 percent in AI-enabled categories. Total sales climbed 39 percent. Prices, as anyone could have predicted, went down, because 500 million of anything tends to do that to the price of one of it.
Meanwhile, the number of non-AI artists on the platform fell 23 percent. Not "stayed flat while adapting to new realities." Fell. Nearly a quarter of them, gone.
Even now, the paper is careful to note that quality held up or improved and that lower-performing human artists were the likeliest to leave, while stronger ones "may have stayed or adapted." One presumes that is meant as reassurance. It reads more like a eulogy with a silver lining stapled to it.
What the Economists Actually Said
Goldberg, in comments accompanying the study: "We're in this transition in the economy between human-centered production" and toward whatever comes next, which is doing a lot of work for one sentence fragment. Elsewhere, the more direct version: "Our results show GenAI is likely to crowd out non-GenAI firms."
On the prospect of AI simply dominating creative markets outright, Goldberg called that "a real policy concern." Sources confirm this is Stanford Business School for "yikes."
The Part Where Nobody Asked the Artists
The paper does not call for banning AI-generated art from creative marketplaces, and it isn't really the study's job to. It documents what the market did once given the option, and the market — as it does — optimized for cheaper and more, with quality-adjusted satisfaction humming along nicely for the people buying stock photography of laptops and handshakes.
The 23 percent of artists who left were not, as far as the record shows, consulted on whether this was an acceptable trade. They found out the way people usually find out: the buyers stopped calling.
And yet — as if this were not enough — the study's real contribution may be less "AI hurts artists," which nobody needed a working paper to establish, and more the specific shape of the hurt: not obliteration, just a slow, well-documented crowding-out, 78 percent more images at a time.
