Artificial intelligence identified a Renaissance masterpiece. Art historians are skeptical.

Exhibiting a painting by the Renaissance master Raphael has generally been reserved for the world’s most famous art venues, such as the Louvre in Paris, the Metropolitan Museum of Art in New York and the Vatican Museums.

But this week, a gallery in the northern English city of Bradford put on display what it says artificial intelligence has identified as a work by the Italian Renaissance painter often mentioned alongside Michelangelo and Leonardo da Vinci. Researchers hope their experimental use of AI will put to rest a decades-long debate about the origins of the painting, known as the de Brécy Tondo, allowing it acceptance alongside Raphael works hanging in cities better known for their art halls.

The two-month exhibition in the Cartwright Hall Art Gallery shows how AI models could upend the small world of professional art attribution, traditionally the province of experts who can devote their entire career to the study of a single artist or period.

Researchers at the Universities of Nottingham and Bradford said in January that an AI-powered facial recognition model showed a 97 percent similarity between the Virgin Mary depicted in the disputed de Brécy Tondo painting and Raphael’s confirmed Sistine Madonna, as well as an 86 percent similarity between the child and the altarpiece’s Jesus. Some experts on the artist disagree.

Hassan Ugail, a professor of visual computing at the University of Bradford who developed the model, said the case shows it’s time for art connoisseurs to modernize their tool kit.

“It has been quite a learning curve to understand the art world and how little they use scientific evidence,” Ugail said. The model researchers used zeroes in on “dimensions the human eye can’t see,” he said.

Art attribution can carry enormous financial stakes, said Richard Polsky, who runs an authentication firm focused on 20th century American artists.

“Say there’s a very small number of genuine works out in the museums. You add another one to the market and can put a tremendous price on it because it’s been in private hands for all these centuries,” Polsky said. “It doesn’t take much for someone with money to say, ‘I want that for my new museum and I’ll pay $100 million.’”

Some artists have a definitive catalogue of attributed works. Others don’t, meaning scholars and the art market must reach consensus on potential new pieces that could fetch millions and change viewers’ understanding of an artist.

Advocates of using AI in the process say it will make the attribution process fairer.

“The reliance on the judgment of a single human expert can be risky due to the potential for human error, subjectivity or biases,” said Carina Popovici, the chief executive of AI authentication firm Art Recognition, who was not involved in the de Brécy Tondo study. She said her group’s algorithm accurately sniffs out paintings made by forger Wolfgang Beltracchi, who admitted in 2012 to faking works by roughly 50 artists.

But in the case of the de Brécy Tondo, experts on Raphael and Renaissance art say they’re skeptical of the AI model’s results.

A private collector bought the painting in 1981, noticing its similarity to the Sistine Madonna, and transferred it to a trust in 1995 for experts to study. Proponents call it a Raphael original. Others say it’s a copy made anytime between the artist’s lifetime and the 19th century.

Rudolf Hiller von Gaertringen, an art historian at the University of Leipzig who co-wrote a compendium of Raphael’s work, said he does not think the painter would have produced a copy of figures depicted in the Sistine Madonna at the height of his career in the 1510s, preferring to vary his subjects.

Patricia Emison, an Italian Renaissance expert at the University of New Hampshire, agreed.

“To repaint the Madonna and child motif that one sees in the [Sistine Madonna] altarpiece is beneath his artistic dignity,” Emison said. “He’s not just looking for the extra buck.”

Polsky and Emison both said they are unsure if an AI model could detect nuances experts see when they look at a work of art.

“If you’re deeply immersed in an artist’s work, you’ve read everything about them. You’ve been to all the museums all over the world to see originals, you’ve been to gallery exhibitions, maybe you’ve owned a few or bought and sold them,” Polsky said. “I don’t think that sort of thing can be taught.”

Despite skepticism from traditional art experts, AI is fast becoming a fixture of the art world.

In addition to distinguishing fake and real artwork, AI-driven models are generating new art and controversy. A group of artists sued several AI companies this year, saying they violated copyright law when using online art as training fodder.

Unlike the lawsuit, a definitive decision on the de Brécy Tondo’s provenance may be hard to reach.

“Attribution is delicate. It’s not something that one practices with absolute certainty,” Emison said.