TruaceTracing the truth around AIMonday, August 17, 2026
Media & Arts·The Trace·Dual reading·Published 2026-08-09

AI brushstroke authentication of Jan van Eyck paintings

Source article: AI analysis casts doubt on Van Eyck paintings in Italian and US museums

Abstract: An analysis of two paintings in museums in the US and Italy by the 15th-century Flemish artist Jan van Eyck has raised a profound question: what if neither were by Van Eyck? Saint Francis of Assisi Receiving the Stigmata, the name given to near-identical unsigned paintings hanging in the Philadelphia Museum of Art and the Royal Museums of Turin, represent two of the small number of surviving works by one of western art’s greatest masters, revered for his naturalistic portraits and religious subjects. The only pr…

TRV-2026-0709JournalismPermanent record — cite & verify
Trace impact reading

Negative state: both sides are scored from claims and sources, not community votes.

P 58The P score combines the specificity and measured human impact of the grounded problem claim with the strength of this Trace’s cited sources.G 53The G score combines the specificity and measured human impact of the grounded gain claim with the strength of this Trace’s cited sources.
AI analysis casts doubt on Van Eyck paintings in Italian and US museums
The quick read

On 7 February 2026, The Guardian reported an AI analysis by Swiss company Art Recognition of two near-identical paintings titled Saint Francis of Assisi Receiving the Stigmata held in Philadelphia and Turin and attributed to Jan van Eyck. The system found no Van Eyck brushstrokes, rating the works 91% negative and 86% negative, while rating The Arnolfini Portrait in London 89% likely authentic.

The finding matters because attribution underpins museum display, scholarship, and public trust in Old Master collections, and a negative AI result directly challenges long-held attributions. Uncertainty remains because condition and restoration history may affect brushstroke analysis, and scholars frame the works as possible studio paintings rather than outright forgeries, leaving open whether a lost autograph original exists.

Main points
  • Art Recognition, a Swiss company that collaborates on research with Tilburg University, conducted AI brushstroke analysis on two near-identical unsigned paintings titled Saint Francis of Assisi Receiving the Stigmata.
  • The Philadelphia Museum of Art version was rated 91% negative and the Royal Museums of Turin version 86% negative for Van Eyck's brushstrokes.
  • The Arnolfini Portrait in the National Gallery in London was rated 89% likely to be authentic in contrast.
  • Scholar Till-Holger Borchert said findings support view that both Saint Francis versions were studio paintings produced in the artist's workshop.
Gain

AI brushstroke analysis by Art Recognition can distinguish authentic Van Eyck work, rating The Arnolfini Portrait as likely authentic.

Problem

AI analysis of two museum paintings attributed to Jan van Eyck found no detectable Van Eyck brushstrokes, casting doubt on their attribution and creating reputational risk for the holding museums.

The rundown

Art Recognition analyzed the Philadelphia and Turin versions of Saint Francis of Assisi Receiving the Stigmata, described as near-identical unsigned paintings among the small number of surviving works by Jan van Eyck. The company reported 91% negative for Philadelphia and 86% negative for Turin, and said it was unable to detect any of Van Eyck's brushstrokes.

CEO Dr Carina Popovici said she expected one painting to be positive and the other negative, but both came out negative, calling the high negative percentages particularly dramatic. Art historian Dr Noah Charney said deeper tests were conducted to confirm the surprising results and suggested both pictures are studio works, implying a lost original more fully by Van Eyck's hand may exist.

What this doesn’t fix

Critics note painting condition and later restorations may affect AI-based brushstroke analysis, leaving uncertainty about whether negative results reflect authorship or altered surfaces.

Sources

Reader signal

How should this claim be treated?

The debate