AI Study Suggests El Greco Painted More of “The Baptism of Christ” Than Scholars Thought
A machine-learning model trained to read paint surfaces at microscopic scale is reshaping a familiar attribution debate around El Greco’s “The Baptism of Christ.” Researchers at Western Reserve University in Ohio say their analysis suggests the artist painted most of the work himself, even though historians have long believed that much of it was completed by his son, Jorge Manuel, and workshop assistants.
The findings, published in Science Advances on April 17, come from a model called Patch, short for pairwise assignment training for classifying heterogeneity. Rather than relying on traditional visual comparison alone, Patch scans a painting in high-resolution 3D, mapping the peaks and hollows left by brushwork and comparing those textures centimeter by centimeter. If the system struggles to separate two areas, the researchers infer that the same hand likely made them.
Before turning to El Greco, the team trained Patch on 25 paintings by nine student artists. The model performed strongly in that test phase, then correctly identified “Christ on the Cross” as the work of a single artist. When applied to “The Baptism of Christ,” however, it produced a more surprising result: it found underlying connections between sections previously thought to belong to Jorge Manuel or other workshop painters.
That does not erase the possibility of later intervention. The lower portion of the painting still appears to have been handled by other hands. But the researchers argue that the differences long noted by art historians could also reflect El Greco experimenting with style, changing brushes, or the physical effects of age on his brushwork.
The painting itself has a complicated history. “The Baptism of Christ” was fixed as the altarpiece in the church of Hospital Tavera in Toledo in 1624, and its authorship has remained a subject of debate ever since. The new study does not settle the question definitively, but it does suggest that the old division between master and workshop may be more porous than previously assumed.
Andrew Van Horn, the study’s lead author, said more work is needed before the model can conclusively identify what he called artistic practice regimes. Still, the method points toward a broader shift in connoisseurship: one in which AI may help reconstruct how Renaissance workshops actually functioned, and how an artist’s hand can be traced across apprenticeship and production.
