A digital creation by ©Creative Desk, tested on curiosity and strong wifi
In 2018, Christie’s stunned the art world by selling a portrait for $432,500. The sitter, Edmond de Belamy, never existed: he was generated by an algorithm trained on 15,000 historical portraits. Signed with the formula that produced him, Belamy symbolised a new era. “For me, that sale was a watershed moment,” says Filip Vermeylen. “AI suddenly moved from being a curiosity to a force the art market could no longer ignore. It impacts how art is created, how it’s valued, and how it’s traded.”
Auctions of the Future, Outrage of the Present
Belamy was only the beginning. In 2024, Sotheby’s sold AI God: Portrait of Alan Turing, painted by a humanoid robot, for $1.08 million. This year, Christie’s mounted its first AI-focused sale, bringing in $728,000 with an 82% sell-through rate. Half the bidders were millennials or Gen Z. Yet backlash was swift: more than 6,000 artists signed a petition accusing AI of exploiting human creativity. Vermeylen pointed out: “Many works in that auction were by well-known artists using AI as a tool. The real question is whether the human is still the artist—or if the algorithm itself has creative value.”
From Tools to Co-Creators
That question starts in the studio. Text-to-image models now allow anyone with a prompt to generate compelling images. “Who is the artist if the work comes from a machine?” Vermeylen asks. “And what does that mean for copyright?” Law offers few answers, often leaving creators unprotected. He acknowledges the appeal: “You could say AI democratises the art world. But it’s a double-edged sword. Established names often capture most of the benefits, while newcomers face uncertain returns.”
One example brings the issue to life. Vermeylen once asked DALL-E to create a portrait of Nelson Mandela in the style of Marlene Dumas. The result echoed her signature look—and even produced a fake signature. “The result looked uncannily like a Dumas, complete with a forged signature,” Vermeylen says. “AI can imitate an artist’s style so closely that, if sold, it could be mistaken for the real thing.”
Garbage In, Garbage Out
AI isn’t only making art; it’s evaluating it. Algorithms can scan brushstrokes, mine auction data, and track public sentiment. Combined with blockchain, they promise greater transparency. Still, Vermeylen cautions: “Garbage in, garbage out. If the data is flawed, the results are flawed.”
“Bias is a particular concern: if certain artists aren’t in the data, the system
will undervalue them,” Vermeylen warns. “AI should support human judgment, not replace it.”
Research on the Edge
For researchers, AI is both tool and topic. At Erasmus University, cultural economists are using machine learning to explore vast catalogues and refine price models. But Vermeylen sees big gaps: “Ethical concerns about biased AI valuations remain speculative. We need real data. In other words, there’s plenty of work to be done for cultural economists.”
Why the Story Still Rules
For all the disruption, Vermeylen stresses continuity. “The art market is still a social system built on networks, trust, and stories,” he says. Museums consecrate artists, collectors rely on intermediaries, and narratives shape value.
“Storytelling has always been central. It’s not art until somebody says so.” And as a historian, he wonders: will AI-generated works one day hang beside Rembrandt, Vermeer or Frida Kahlo—or will they be dismissed as what he calls “AI slop”? For now, he says, “the jury is still out.”
As Professor of Global Art Markets at Erasmus University Rotterdam, Filip researches the history and functioning of art markets worldwide, co-founded the Rotterdam Arts & Sciences Lab (RASL), serves on the board of TIAMSA, and is contributing an exhibition on the history of the art market in Vienna which will open in January, 2026.
A digital creation by ©Creative Desk, tested on curiosity and strong wifi