AI does not confirm Raphael’s authorship of part of the painting "Madonna with the Rose"

The inconsistent style of depiction of St. John’s face in the Madonna of the Rose has been a subject of debate among art historians for years. Artificial intelligence may have cracked the code to a long-standing question about Raphael’s famous painting that has puzzled the art world for years.

Part of Raphael’s "Madonna of the Rose" may not have been painted by the Italian Renaissance master, according to findings published in the journal Science Heritage in December 2023 that are now surfacing and circulating online.

Raffaello Sanzio da Urbino is considered one of the most famous Renaissance artists for his paintings and frescoes, including the School of Athens in the Vatican. However, the Madonna of the Rose, on display at the Prado National Museum in Madrid, has long been the subject of controversy as the image of St. John seems out of keeping with Raphael’s style and less refined than other elements of the painting. The painting depicts the Madonna with the infant Jesus, as well as St. Joseph and St. John.

Scientists from the University of Nottingham, the University of Bradford and Stanford developed a special neural network algorithm to compare parts of the painting with other works by Raphael. The team used photographs of verified Raphael paintings to “train a computer to recognise his style down to the smallest detail, from brushstrokes, colour palette, tones and every aspect of the work,” says Hassan Ugail, a computer scientist at the University of Bradford.

“When we tested the painting as a whole, the results were inconclusive,” Ugail said in a statement. “We then tested individual parts, and while the rest of the painting was confirmed as Raphael, Joseph’s face was most likely not Raphael.”

In a press release, the scientists said it was likely that the face of St. John was painted by one of Raphael’s students, Giulio Romano. While the study’s findings are not 100 percent conclusive, Ugail noted that the computer model examines artworks at a microscopic level "with 98 percent accuracy."

Christopher Brook, an honorary research fellow at the University of Nottingham who was also involved in the project, said in a press release that the analysis “promises to be a useful complementary tool in future studies of this kind, alongside well-established techniques such as spectroscopy.”