The art world is grappling with a potentially seismic shift in understanding the work of Jan van Eyck, the 15th-century Flemish master. New analysis, utilizing artificial intelligence, is casting doubt on the attribution of two paintings long considered to be by his hand: near-identical depictions of Saint Francis of Assisi Receiving the Stigmata, currently housed in the Philadelphia Museum of Art and the Royal Museums of Turin.
The findings, conducted by Art Recognition, a Swiss firm collaborating with Tilburg University in the Netherlands, suggest that neither painting exhibits the brushstrokes characteristic of Van Eyck. The Philadelphia version registered a “91% negative” result, while the Turin painting came back “86% negative.” This doesn’t necessarily mean the paintings are forgeries, but rather raises the possibility that they are studio works – created by assistants in Van Eyck’s workshop, but not directly by the artist himself.
The implications are significant. Van Eyck is revered as a pivotal figure in the history of art, particularly for his mastery of oil painting and his groundbreaking realism. His portraits, like the famed Arnolfini Portrait, are celebrated for their lifelike detail and innovative use of light and shadow. With a relatively small number of universally accepted works attributed to him – fewer than 20, according to some estimates – each painting carries immense weight.
Till-Holger Borchert, director of the Suermondt-Ludwig-Museum in Aachen and a leading Van Eyck scholar, acknowledged being “surprised” by the analysis but emphasized the need for further investigation. He noted that the findings align with existing scholarly debate suggesting that some works attributed to Van Eyck may, in fact, be products of his studio. This was a common practice in the 15th century, where workshops often collaborated on paintings, with the master overseeing the process but not necessarily executing every detail.
The AI-driven analysis isn’t simply identifying a lack of Van Eyck’s brushwork; it’s suggesting a fundamental re-evaluation of how we understand artistic creation in the early Renaissance. As Dr. Noah Charney, an art historian, explained on his podcast, the romanticized image of the “lone artist” toiling in isolation is a relatively modern construct. Workshops were collaborative environments and attributing authorship can be far more complex than simply identifying who physically applied the paint to the canvas.
Carina Popovici, Art Recognition’s chief executive, expressed surprise at the consistency of the negative results. “I expected that, if one painting was negative, the other would be positive. But no, both came out negative,” she told The Guardian. She acknowledged the potential for a difficult conversation with the Philadelphia and Turin museums, adding, “I’m guessing that the Philadelphia and Turin museums won’t be happy. It’s not good news on these paintings.” Both museums have been contacted for comment.
This isn’t the first time Art Recognition’s AI analysis has challenged established attributions. In 2021, their technology indicated that Peter Paul Rubens’ Samson and Delilah in the National Gallery was “91% negative,” supporting long-held doubts about its authenticity. More recently, in 2024, the company identified 40 fake paintings being offered for sale on eBay.
The use of AI in art authentication is a rapidly evolving field. While the technology isn’t foolproof – critics point to the potential for factors like a painting’s condition and subsequent restorations to influence the results – it offers a powerful new tool for art historians and museums. The ability to analyze brushstrokes with a level of precision previously unattainable could lead to a more nuanced understanding of artistic practice and attribution.
The debate surrounding the Saint Francis of Assisi Receiving the Stigmata paintings comes at a particularly relevant moment. The National Gallery in London is preparing to host a major exhibition of Van Eyck’s portraits in November, an event billed as a cultural highlight of the year. The exhibition promises to showcase the artist’s remarkable skill and innovation, but the recent AI analysis serves as a reminder that even the most celebrated masterpieces are subject to ongoing scrutiny and reinterpretation.
Charney suggests that the negative results for both paintings could indicate the existence of a lost original by Van Eyck, with the Turin and Philadelphia versions being copies created within his workshop. This possibility underscores the enduring mystery surrounding Van Eyck’s oeuvre and the challenges of definitively attributing authorship to works created centuries ago. The question isn’t simply whether these paintings are “by” Van Eyck, but rather what they tell us about the collaborative nature of artistic production in the 15th century and the enduring legacy of one of art history’s most influential figures.
