Some artworks arrive with a whisper. Others arrive with an argument already attached. Una Vilana Windisch belongs firmly to the latter category. From the moment this quiet, piercing portrait of a peasant woman resurfaced, it seemed determined to provoke disagreement, scrutiny, and, ultimately, a rethinking of how we decide what is, and is not, an authentic work by one of the most scrutinized artists in Western art.

In 2023, Art Recognition was approached by a European private collector with a delicate portrait on vellum, dated 1505 and attributed to Albrecht Dürer. The sitter is an unidealized rural woman (likely from today’s Slovenia, as Windisch was the German name for Slovenians), rendered with extraordinary economy and psychological presence. The drawing’s significance was immediately apparent, not least because a nearly identical version of the same composition, executed on paper and firmly attributed to Dürer, has long been held in the collection of the British Museum.

The existence of multiple versions of a composition is nothing unusual in Dürer’s oeuvre. He was famously methodical, producing preparatory studies, variants, and refinements as part of his working process. What made Una Vilana Windisch exceptional was not merely its similarity to the British Museum drawing, but the degree of that similarity. When the two works were digitally superimposed, their contours aligned with uncanny precision. This was not a loose copy, nor a workshop echo. It was, line for line, nearly the same drawing. 

The portrait’s modern history is equally dense. Since the 1970s, Una Vilana Windisch has undergone sustained art historical, forensic, and technical scrutiny. A handwriting expert compared the monogram, date, and inscription with authenticated letters by Dürer, concluding that the writing was a genuine, spontaneous autograph rather than a traced or imitative signature. Scientific analysis reinforced this conclusion. Infrared imaging revealed an underdrawing consistent with Dürer’s working methods. Ultraviolet light showed no evidence of modern retouching. Material testing confirmed the use of iron-gall ink on lime-treated vellum, entirely consistent with early sixteenth-century practice.

And yet, despite this formidable body of evidence, consensus proved elusive. Several respected scholars supported the attribution, including Theodor Musper and Herbert Peter, who included the drawing in the 1989 Berlin exhibition “Dasein und Vision”. Others remained unconvinced. Most notably, John Rowlands rejected the vellum version outright, maintaining that only the paper drawing in the British Museum could be accepted as authentic. The result was a familiar stalemate: strong evidence on both sides, and no mechanism capable of resolving the impasse definitively.

IIt was at this point that the collector turned to Art Recognition for an additional, independent layer of analysis. Rather than replacing traditional connoisseurship, the goal was to subject the drawing to a different kind of scrutiny altogether. Using a dataset of 144 confirmed works by Dürer across ink, chalk, and charcoal, alongside an equal number of forgeries, imitations, and synthetic images, the AI was trained to recognize the artist’s distinctive stylistic fingerprint at a granular, statistical level.

The results were striking. Una Vilana Windisch was classified as authentic with an 82.2 percent probability. Even more intriguingly, the British Museum’s paper version, long regarded as the unquestioned original, was also analyzed and received a slightly lower authenticity score. This unexpected reversal did not invalidate the museum drawing, but it complicated the story in a way that purely traditional methods had not. The findings attracted international attention and were reported in The Art Newspaper, bringing the debate into the public eye once again.

The Una Vilana Windisch case is not about AI delivering a final verdict from on high. It is about adding a new voice to an already crowded conversation. In a field where there is no universal standard for authentication, and where reputations, markets, and institutions often pull in different directions, AI offers something both modest and powerful: a transparent, data-driven perspective that can support, challenge, or refine existing judgments. In this case, it did not silence disagreement. It made the discussion sharper, richer, and harder to ignore.

< Back