AI : The world of art authentication rarely grabs mainstream headlines, but last year, it did just that. A group of UK researchers announced a groundbreaking claim: an anonymous, centuries-old painting known as the de Brécy Tondo was likely the work of Renaissance master Raphael. The bold assertion, with potentially massive financial repercussions, wasn’t the only reason it made news. The method used to reach this conclusion was what truly caught people’s attention—artificial intelligence.
The research team, led by Professors Christopher Brooke of the University of Nottingham and Hassan Ugail of the University of Bradford, utilized a facial recognition model to analyze the Madonna in the de Brécy Tondo. They compared it to other portraits of the Madonna and found a 97% match with Raphael’s Sistine Madonna altarpiece. This led the team to conclude, as Ugail stated, that “identical models were used for both paintings and they are undoubtedly by the same artist.”
However, the excitement surrounding this discovery was short-lived. Eight months later, Art Recognition, a Swiss AI company, challenged the claim with its own AI model, determining with 85% certainty that the de Brécy Tondo was not a Raphael creation. In an op-ed, Carina Popovici, Art Recognition’s founder, defended her company’s findings, highlighting the expertise of her team, which included several art historians, and the sophistication of their model, trained on authentic and forged Raphael paintings. However, she was careful not to discredit Brooke and Ugail. “The most straightforward explanation for the strong discrepancy between the two results is that the models are essentially addressing different questions,” Popovici explained.
The controversy, dubbed the “battle of the AIs” by the Guardian, did little to sway skeptics of AI’s role in art interpretation. Instead, it underscored the ongoing debates within the art world as AI technology increasingly integrates into its most revered institutions.
While AI is already curating exhibitions and biennials, could it also reshape how we study and perceive historical art?
The intersection of AI and art history is not merely a technological issue but also an ideological one. In 2013, scholar Johanna Drucker published “Is There a ‘Digital’ Art History?,” a paper exploring the implications of digital tools in art history. Drucker concluded that while digital technology has made art history more accessible, it hasn’t altered the field’s core approaches or methodologies.
Drucker’s paper, along with a subsequent essay by art historian Claire Bishop, sparked debate within the digital humanities—a field that combines advanced computational techniques with the study of humanities disciplines like art and literature. Bishop’s essay, “Against Digital Art History,” argued that the push toward digital methods reflects a broader socioeconomic issue: the pressure to quantify and optimize knowledge, driven by neoliberal ideals.
“Digital art history, as the belated tail end of the digital humanities, signals a change in the character of knowledge and learning,” Bishop wrote.
She questioned whether the empirical methods of computational analysis could truly enhance the theoretical interpretations central to the humanities.
Despite the rapid advancements in AI since these essays were published, the underlying concerns raised by Bishop and Drucker remain relevant. Their critiques arrived during a time when funding for STEM fields was often prioritized over the humanities, a trend that continues to influence the discourse around AI in art history.
Some experts, like Amanda Wasielewski, a digital humanities professor at Uppsala University in Sweden, are less concerned about AI replacing art historians. “I don’t think there’s any AI or machine learning technique that could ever replace an art historian,” she stated. Wasielewski acknowledges the usefulness of AI in certain aspects of art history, such as archival management, but remains cautious about its potential to reintroduce outdated methodologies.
In her book Computational Formalism, Wasielewski explores how machine learning has revived the strict, formalist methodologies of art study that dominated the early 20th century.
This approach, which focuses on an artwork’s physical properties rather than its contextual background, fell out of favor during the cultural shifts of the 1960s and ’70s. However, with the rise of AI, Wasielewski warns that this dogma could resurface.
Machine learning models excel at formal analysis and pattern recognition, making them valuable tools in art authentication and archival work. But as we increasingly rely on these systems, Wasielewski suggests we risk neglecting the critical frameworks that have evolved over decades. “When you think that somehow you’re going to draw out objective things from a formalist methodology,” she noted, “you don’t do any extra methodological work.”
The resurgence of formalist thinking through AI does not necessarily mean a return to old practices. The concept of “distant reading” in literature, introduced by literary historian Franco Moretti, has been adapted to “distant viewing” in art. This method analyzes large datasets to identify trends and patterns across time, place, and style. Researchers like Leonardo Impett and Fabian Offert have used this approach to draw new connections between historical and contemporary art.
Their paper, titled There Is a Digital Art History, highlights the potential of AI to enhance art historical work—though with the caveat that these models are only as good as the data they are trained on. “We can have the benefits of these new models, but at the same time,” Offert stated, “we have to always critique [them] and figure out the limitations of their…weird machinic visual culture.”
Wasielewski agrees, emphasizing that these AI models are not autonomous entities but tools created by humans, with all the biases that entails.
“They’re not magical machines,” she said. “We need to question not just how these tools are applied, but where they have their origins, what kind of data they were trained on, what biases might be contained within [them].”
The debate over AI in art history is far from over. Even as AI continues to challenge traditional methods, it also opens new avenues for dialogue between technologists and historians. Popovici’s collaboration with German art history professor Nils Büttner on a research project exemplifies this potential. Despite their different approaches—Büttner’s traditionalist methods and Popovici’s AI-driven analysis—they reached similar conclusions about a painting attributed to Anthony van Dyck, demonstrating the value of combining old and new methodologies.
As Wasielewski pointed out, the conversation about AI’s role in art history is just beginning. “We in the field of art and in the field of art history need to be involved in these conversations,” she said. As AI continues to evolve, so too will the debates surrounding its place in the study and appreciation of art.