When Artificial Intelligence Becomes a Provenance Tool

A centuries-old attribution dispute surrounding a major El Greco altarpiece may finally have a credible answer — and the instrument that cracked it open was not a conservator's scalpel or an archival discovery, but a machine-learning algorithm trained on brushstroke data. Published in the peer-reviewed journal Science Advances, the study deployed A.I. image analysis to determine whether the Adoration of the Shepherds, long attributed in part to El Greco's son Jorge Manuel Theotokópoulos, was in fact the master's own work. The findings point firmly toward Domenikos Theotokópoulos — El Greco himself — and the implications stretch well beyond one Spanish altarpiece. For serious collectors of Old Masters, and for the growing cohort of Asian buyers acquiring European works at auction, this development is a landmark moment in provenance science.

The Painting, the Dispute, and the Data

The altarpiece in question was completed around 1612–1614 for the Santo Domingo el Antiguo monastery in Toledo, Spain, where El Greco had established his reputation decades earlier. For generations, art historians debated whether Jorge Manuel — who frequently assisted his father and continued painting after El Greco's death in 1614 — had executed significant portions of the work. The attribution uncertainty suppressed the painting's scholarly standing and, by extension, its theoretical market ceiling. Works with contested authorship routinely sell at discounts of 30–60% against fully authenticated pieces by the same artist. A confirmed El Greco attribution at auction, by contrast, commands extraordinary sums: his Saints Peter and Paul achieved €4.2 million at Sotheby's, while authenticated panel works have cleared the €10 million threshold at major European houses.

The research team trained their neural network on a corpus of definitively authenticated El Greco canvases, teaching the model to recognise micro-level stylistic signatures — impasto texture, directional brushwork, tonal gradation — that are invisible to the naked eye but statistically consistent across a single artist's hand. When the algorithm was applied to the disputed altarpiece, it returned a match confidence aligned with El Greco's authenticated output, not his son's. The methodology mirrors techniques already used in fingerprint analysis and medical imaging, now repurposed for art authentication with rigorous academic oversight.

Why This Matters to Asian Collectors

Asian collectors — particularly those based in Hong Kong, Singapore, Tokyo, and increasingly mainland China — have accelerated their acquisition of Western Old Masters over the past decade. Christie's Hong Kong and Sotheby's Asia have both reported sustained demand for works with deep European provenance, with buyers drawn to the combination of historical gravitas and relative scarcity. A fully authenticated El Greco commands a place in the rarest tier of Western art: fewer than 300 works are accepted in the authoritative catalogue raisonné, making each confirmed attribution a significant market event. For collectors building cross-cultural holdings — pairing, say, a Song dynasty ceramic with a Mannerist canvas — authentication certainty is not a footnote but a financial foundation.

The broader significance lies in what A.I.-assisted attribution signals for the market going forward. Christie's, Sotheby's, and specialist authentication bodies have all begun exploring machine-learning tools as supplements to traditional connoisseurship. For buyers in Asia, where the secondary market for Western art is still maturing and due diligence infrastructure is less entrenched than in London or New York, algorithmic authentication offers an additional layer of confidence. It also raises the stakes for works currently sitting in private Asian collections with incomplete documentation — a not-uncommon situation given the volume of European art that moved through Hong Kong and Singapore dealers in the 1970s and 1980s.

Collection-Building Insight

The El Greco case is a reminder that provenance is not static. A work dismissed or discounted for decades can be reappraised upward when new analytical tools emerge, and collectors who acquire undervalued pieces with credible — if contested — attributions stand to benefit substantially. The premium between a disputed attribution and a confirmed one can represent millions at the top of the market, but even at the mid-market level, a clean authentication can move a work from the €80,000–120,000 range into the €300,000-plus bracket. For Asian collectors with a long time horizon and an appetite for research-driven acquisition, Old Masters with unresolved attribution questions represent one of the few remaining areas where patient, informed buying can still generate asymmetric returns. The tools to resolve those questions are now, demonstrably, more powerful than ever.

• Artist: El Greco (Domenikos Theotokópoulos), 1541–1614
• Work in question: Adoration of the Shepherds, c. 1612–1614
• Authentication method: Neural network trained on confirmed El Greco corpus, published in Science Advances
• Market benchmark: Authenticated El Greco works have achieved €4.2M–€10M+ at major auction houses
• Attribution discount: Disputed works typically sell at 30–60% below confirmed attributions
• Catalogue raisonné: Fewer than 300 accepted works, making each confirmation a significant scarcity event

🥃 Building a whisky cask collection? Whisky Cask Club curates rare Scottish casks for private collectors across Asia.