The clue was a photograph of a photograph. Sometime in 1960, the Catalan art historian José Gudiol Ricard stood in front of a small Gothic panel and made a black-and-white negative of it — the kind of careful documentary image that the great cataloguers of Spanish painting produced by the thousand and then filed away. The panel showed a religious procession winding up a hill. Within a few years it would be gone from Spain, and for the better part of seventy years no one there could say where it had gone. This summer, someone finally asked a machine.
The machine was Google Lens, the reverse-image search built into hundreds of millions of phones. A researcher at the ZamorArte Foundation, a heritage group based in Zamora, scanned Gudiol's old negative, fed the grainy monochrome image into the app, and waited. What came back was not a maybe. "It gave us two exact matches," the foundation's Jaime Gallego Boyano told reporters — "two matches because people who had visited the Springfield museum had taken a photo and uploaded it to the internet." A lost medieval painting had been hiding, it turned out, in other people's holiday snapshots.
The panel that walked out of a church
The work is a panel attributed to Nicolás Francés, and it once belonged to a four-part altarpiece in the church of San Miguel in Villalpando, a small town on the wheat plains of the province of Zamora. Its subject is the Procession of Mount Gargano — a scene from the legend of Saint Michael, whose apparition on a mountain in southern Italy gave the medieval world one of its most enduring cult sites. Four panels, one dedicated church, one archangel: an entirely ordinary piece of provincial devotion, of the sort that fills the storerooms of Spain and rarely troubles the international market.
Nicolás Francés himself was not provincial at all. His very name — "the Frenchman" — records that he arrived from across the Pyrenees, and by the 1430s he was the dominant painter in the city of León, entrusted with work for its great cathedral. He is now counted among the leading masters of the International Gothic in the Kingdom of León, that late-medieval style of gilded grounds, courtly grace and jewel-bright colour that spread across Europe before the Renaissance rewrote the rules of looking. The gold-ground panels of his workshop are exactly the kind of object modern eyes find easy to underrate and easy, therefore, to lose track of.
A lost medieval painting had been hiding, it turned out, in other people's holiday snapshots.
At some point after Gudiol photographed it, the Procession left Villalpando. The panel vanished from local knowledge in 1957, swept up in the quiet, decades-long dispersal of Spanish church art through dealers and private hands — the same undocumented current that carries so many objects out of their home countries and into the collections of the wider world. On the Spanish side, the trail simply went cold.
Two matches on a museum wall
Where it went cold was, as it happens, a well-lit gallery in New England. The panel had been acquired in 1964 by the Michele and Donald D'Amour Museum of Fine Arts in Springfield, Massachusetts — one of the cluster of institutions known collectively as the Springfield Museums. There it was hung, catalogued and, by every account, correctly identified as a Nicolás Francés. It was not buried in a basement or mislabelled as a copy. It was on view, published in the museum's own literature, seen by every visitor who walked that particular room.
And that is the quietly modern twist in the story. The reverse-image search did not defeat a disguise. It defeated a gap — the distance between a Spanish researcher who knew the panel by a seventy-year-old negative and an American museum that knew it by an accession number, two accurate records that had never been introduced to each other. What Google Lens supplied was the introduction, and it did so through the most banal artefact of contemporary travel: the phone photographs that ordinary tourists take in front of paintings and upload without a second thought. Somewhere in that ocean of casual images were two frames good enough for an algorithm to say, with something close to certainty, this is the same object.
Why a machine could see it
It is worth being precise about what actually happened, because the headline version — "AI finds lost masterpiece" — flattens something more interesting. Reverse-image search of this kind does not understand Saint Michael, or Gothic drapery, or the province of Zamora. It matches visual patterns: the silhouette of a hill, the arrangement of figures, the particular geometry of a gilded arch. Feed it a 1960 monochrome negative and a 2020s colour snapshot of the same panel, and the underlying shapes line up closely enough to trip a match, across the gulf of film grain, lighting and six decades.
What the technology cannot do is the part that still belongs entirely to people. A machine can propose that two images show the same thing; it cannot confirm that the object on a Springfield wall is the panel that left Villalpando, rather than a sibling, a copy, or a coincidence of composition. That confirmation is the slow, human labour of provenance — matching dimensions, tracing the dealers who handled the picture in the later twentieth century, reading the museum's acquisition file against the Spanish parish records. The foundation has been careful to frame its find as the opening of that inquiry, not the end of it, and it is now in contact with the Springfield museum to compare notes. The reverse-image hit is a lead. The case is built the old way.
An honest word about "lost" and "found"
There is a temptation, with a story this neat, to declare a rescue. It is worth resisting. Nothing has changed hands; no wrong has yet been righted; the panel remains, as far as anyone has said, the lawful property of an American museum that has cared for it in good faith for sixty years. It was never truly lost in the sense of being hidden — it was lost only in the sense that two communities had lost sight of each other. Whether the discovery leads anywhere beyond a shared spreadsheet — a loan, a joint exhibition, a conversation about where the panel ultimately belongs — is unknown, and the responsible version of the story leaves that door open rather than pretending to walk through it. ZamorArte has spoken of hoping to show the Villalpando panels together one day; that is an aspiration, not an event.
What is genuinely new here is the method, and it deserves neither hype nor dismissal. For a century, the reuniting of scattered altarpieces has depended on a scholar happening to remember the right photograph in the right archive at the right moment — a system that works, but only as fast as human memory and luck allow. A phone app that can put a 1960 negative against the entire uploaded photographic record of the tourist age changes the odds. It will not authenticate anything; it will not settle a single question of ownership. But it can do the one thing that used to be hardest of all: notice that a picture on a wall in Massachusetts and a picture in a Spanish filing cabinet are, against all expectation, the same picture. The rest, as ever, is patience.
It is a small object at the centre of all this — one gilded panel from a country church, not a headline masterpiece. But rediscovery, again and again, turns out to be less about buried treasure than about a community re-learning what it once knew, and the tools that let it look again. A lost Constable can hang above a staircase in Texas for fifty years; a lost Sofonisba can wait a full century for the right eye. This time the eye was borrowed from a phone — and it still needed a human being to decide to look.
Frequently asked questions
How did Google Lens find a lost painting? The ZamorArte Foundation scanned a black-and-white negative made in 1960 by the art historian José Gudiol Ricard and ran it through Google Lens. The reverse-image search returned two exact matches — colour photographs that tourists had taken inside the D'Amour Museum in Springfield, Massachusetts, and posted online — placing the panel on a specific museum wall.
What is the Procession of Mount Gargano? A late-Gothic panel, once part of a four-panel altarpiece dedicated to Saint Michael in the church of San Miguel in Villalpando, Zamora. It is attributed to Nicolás Francés, a French-born painter who led the workshop of León between about 1424 and 1468.
Was the painting really lost? Only from Spain's point of view. It went missing from its church in 1957, but the Springfield museum bought it in 1964, catalogued it correctly and kept it on view. The search reconnected two accurate records rather than uncovering a hidden object.