When AI-generated art enters the market, consumers win—and artists lose

In 2024, the New York Times sued OpenAI for copyright infringement, claiming the company had used the newspaper’s content to train its algorithms without permission or payment.

The case, one of several AI-related copyright suits, is another sign of fast-changing times. “We’re in this transition in the economy between human-centered production and a more generative-AI-centered production,” says Samuel Goldberg, an assistant professor of marketing at Stanford Graduate School of Business. “But we train AI models on human outputs, especially in creative industries like art, writing, music, and images. Those human outputs become the inputs.”

A central component of the Times lawsuit, Goldberg says, “ultimately boils down to this question of whether AI output is a substitute for the human inputs it uses.”

That question is also at the center of a new study by Goldberg and H. Tai Lam of the University of California, Los Angeles, which examines what happened when an online marketplace allowed AI-generated art to compete with images made by people. The paper is available on the SSRN preprint server.

“We have some idea of how GenAI might impact productivity, but ultimately, what this productivity means for the market, competition, and demand is an open question. We wanted to see how markets react to the introduction of these types of technologies on both the supply and demand sides,” Goldberg says.

The results were striking: Once generative AI (GenAI) entered the market, the total number of images for sale skyrocketed, while the number of human-generated images fell dramatically. On the flip side, consumers showed a taste for the influx of AI-generated images, choosing GenAI images over human-generated ones.

“Our results show GenAI is likely to crowd out non-GenAI firms and goods,” Goldberg and Lam conclude. Clearly, this is unwelcome news for creators. Yet this brought increased competition, quality, and variety on the platform—benefits for buyers. These findings provide a glimpse into the future of markets for creative goods, including music, novels, and movies.

Image of a changing market

To study the impacts of AI-generated content, Goldberg and Lam worked with data from a large online platform that provides consumers with access to nearly 500 million images and videos. “The buyers might be writers who write blog posts or articles who want images for the piece, or a business that needs an image for the pamphlet,” Goldberg says.

On the other side are the sellers, or the producers of those images—artists, photographers, and illustrators. These were mostly individuals or smaller firms, as Goldberg notes, “Maybe a group of two or three people, but some larger firms that produce tens of thousands of images and have many, many photographers working for them.” The study’s sample included over 3.2 million unique images and 62,000 artists and producers.

Human-generated images were the only goods on the platform until December 2022, when it began allowing images made with generative AI and requiring that they be labeled as such. This change enabled the researchers to “look at the before and after,” as Goldberg explains. “After we see generative AI enter, we see artists making decisions about whether or not to adapt and how to produce their images, if they want to use AI tools or not.”

The platform also prohibited AI images from including any references to brands and from being offered on markets primarily for editorial use. That allowed for a cleaner comparison between markets on the platform where AI was allowed and those where it wasn’t.

The researchers created measures of content and quality of every image studied, partly to understand how similar or different they were. “That way, we can compare AI images to non-AI images and whether the AI images contribute variety to the platform,” Goldberg says.

The introduction of AI images significantly changed the dynamics for sellers and buyers on the site. The researchers saw a “huge increase in images arriving on the platform after the introduction of generative AI,” Goldberg says, 78% more images per month than in markets without AI images. “This was driven almost exclusively by generative AI production.”

The number of producers also grew significantly: There were 88% more active firms per month in markets with GenAI images than in markets without AI images. The increase was driven by sellers using AI and was accompanied by an additional 23% drop in non-AI artists. “We see an exit of non-generative-AI artists from the site,” Goldberg says.

Alongside these shifts, image quality improved overall. “There was also quality improvement specifically among non-AI images,” Goldberg says. “Competition increased, and the non-generative-AI artists who left the platform were probably lower-quality artists. The good ones stay and the bad ones leave. That’s traditionally what you hope competition would do.”

The AI images also added variety to the platform. “They’re not just copies of one another or of existing images,” Goldberg says. “That’s better for consumers.”

A complex picture for creators

Meanwhile, total sales rose by 39% after AI images arrived, but purchases of non-AI images dropped. This implies AI images compete directly with non-AI images. “Our evidence suggests that GenAI images look like substitutes for non-AI images,” Goldberg says.

“Our findings seem to be good news for consumers,” Goldberg says. “There’s more variety on the platform and some quality benefit, and they’re purchasing more.”

“For producers, it’s more tricky,” he continues. Some sellers have turned to generative AI and may be selling more images than before. “But if you were not as competitive before GenAI, you’re being squeezed out of this platform. Is that a bad thing? I’m not sure.”

A more subtle issue is the potential diminishment of human creativity due to the sudden flood of generative AI. “We as a society believe that art created by humans is somehow better than art created by machines,” Goldberg notes. “So one question you might ask is whether we’re at risk of these markets being completely dominated by generative AI, squeezing humans out. That’s a real policy concern.”

Still, he and Lam conclude that their findings do not support banning the use of GenAI in creative goods. Considering its benefits to consumers and the market, the challenge, they say, is “ensuring equitable access to GenAI technologies and non-AI artists.”

And that New York Times lawsuit against OpenAI? This research has some implications for the case (though Goldberg emphasizes he’s not an attorney): “Our paper suggests the generative AI outputs are substitutes for the human-created inputs, and lawyers are arguing that creators of images used to train AI should be compensated for that.”