AI Rewrites Two Royal Identities in Geneva's Liotard Exhibition

A portrait long identified as the child Marie-Antoinette is actually her sister Marie-Caroline — and AI helped prove it. The finding anchors the exhibition Liotard & Marie-Antoinette: Faces Under Investigation, which opened this month at Geneva's Musée d'art et d'histoire (MAH) and runs through February 7, 2027.

The show presents all 11 chalk drawings that Jean-Étienne Liotard (1702–1789) made in 1762 of Empress Maria Theresa of Austria's children, plus a newly acquired portrait of their father, François-Étienne de Lorraine. The portraits of Marie-Antoinette and Marie-Caroline have been retitled to reflect the corrected identities.

How did the mix-up come to light?

The detective story starts with Catriona Seth, a French literature professor at All Souls College, University of Oxford. Twenty years ago, preparing a book on Marie-Antoinette, she put a beloved childhood portrait by Liotard on the cover. Years later she noticed a problem: the young girl wears the chivalric medal of the Order of the Starry Cross. Marie-Antoinette received that medal too — but only after 1762, when Liotard was in Vienna drawing Maria Theresa's many children.

The drawing, long believed to show Marie-Antoinette, was in fact Marie-Caroline. "I love being able to say that to people, because it just shows that research is a continuous process and that we also learn by our mistakes," Seth said. "There are things that can be discovered, even about great portraits, great texts. We have to think about them and question things."

What does AI add to art history?

Proving one swap was only half the job. Seth also suspected the portrait labelled Marie-Caroline was really Marie-Antoinette. She tried a consumer face-aging app on it and compared the result to an adult portrait of the French queen. "It wasn't very useful," she admits. "But I thought: There's an idea here."

That idea became INTERART, a joint experimental project among the MAH (holder of the world's largest Liotard collection), the University of Oxford, the Idiap Research Institute and the University of Lausanne's School of Criminal Justice. The team includes Marie-Eve Celio-Scheurer, the MAH's senior curator and head of graphic arts; David-Olivier Jacquet-Chiffelle, professor of forensic science at Lausanne, home to the world's oldest school of forensic science; and Sébastien Marcel, an expert in biometrics (automated face recognition) and face recognition.

The technical challenge is real. Current facial recognition systems were built for photographs and can already handle differences in pose, lighting, expression, age and image quality. Paintings add another layer: two portraits of the same sitter can differ sharply in style and technique. The models must isolate facial constants from changes introduced by artistic practice. One caveat — a work has to be created directly before the sitter, not copied from another portrait, to be a reliable candidate.

Will algorithms replace the art historian?

No, and the researchers are explicit about that. The system generates thresholds of probability, not verdicts. "There is never a 100 percent correct answer, and this is what also keeps the project interesting," Celio-Scheurer said. "It gives you a potential."

The practical payoff lies with museums' countless portraits of "an unknown woman" or "an unknown man." If databases hold portraits with confirmed sitters, known ages and documented sittings, an algorithm could suggest matches for unidentified works. "It might give us leads," Seth explains — a bit like running a suspect's fingerprints against a criminal database and hoping for a hit.

Traditional methods — provenance, inscriptions, archives, costume and jewelry analysis — stay central. "It's not that AI will replace research in art history; it's more of a support or tool that can give indications," Celio-Scheurer notes. "At the moment it's a support, but who knows how far this tool can be developed?" The project's early results and accompanying catalog now give visitors in Geneva a first look at what that support might become.