Paintings carry more than the artist’s hand. Over a century, varnishes yellow, linings are added, damaged edges are trimmed and well-meaning restorers retouch what they believe is missing. William Merritt Chase’s A Comfortable Corner (around 1888) – a woman in a deep blue kimono reclining in a richly furnished interior, holding a Japanese uchiwa fan – had been through at least two such interventions, most substantially in the 1960s. Now it has been restored again, and this time artificial intelligence played a small but telling role.
An old photograph as key witness
The painting belongs to the Parrish Art Museum in Water Mill on Long Island, which has a deep connection to Chase: the Gilded Age painter spent time on the East End, and the museum holds a large body of his work and family archives. The restored canvas is a centrepiece of the exhibition “William Merritt Chase and Walter Granville-Smith: American Impressionists of the East End”, opening on 14 November 2026 and running until 21 March 2027.
The decisive clue was a black-and-white photograph believed to predate the 1960s restoration. It showed more of the composition than survives today – suggesting the canvas was reduced in size, perhaps to fit a new stretcher. It also showed details that had vanished: a curl of hair beneath the sitter’s ear, parts of a decorative screen, delicate passages later covered by heavier retouching.
What the software did
Conservator Suzanne Siano and her team at Modern Art Conservation used an AI tool to merge the historical photograph with a current colour image into a composite that highlighted where the two diverged. The benefit, she explains, was mainly speed: the comparison was not instant, but much faster than building it by hand in image-editing software.
The composite pointed the team to areas worth a closer look. The most striking find was Chase’s own signature, which had become extremely faint, with some letters apparently overpainted. Because Chase signed in a loose, almost abstract hand, it may have been mistaken for part of the carpet pattern during the earlier treatment.
A tool, not a conservator
What makes this story interesting is the restraint. The team did not simply restore every difference the software flagged. The old photo was low-resolution and not necessarily a record of the original state; the investigation also turned up pentimenti – changes Chase made himself – that complicated the picture. Every finding was weighed against magnified examination, ultraviolet imaging, knowledge of Chase’s materials and comparison with his other works. Some earlier losses, Siano suggests, may have come from restorers cleaning towards what they thought was the original surface while removing the artist’s own additions.
Her verdict is a useful corrective to the hype: conservation, she says, is a combination of science, art history, craft, manual skill and artistry – and “AI is none of those things”. It was the first time her studio used AI on a painting, funded by a Bank of America grant. She expects it to become valuable mainly as a research aid, for example in tracking down historical images.
Why it matters
Museums worldwide are under pressure to show they are innovating. The Parrish case is a good model: technology used transparently, as one source of evidence among many, with human judgement firmly in charge. And visitors on Long Island will now see a Chase that looks a little more like the one he actually painted.