New machine-learning analysis suggests El Greco painted most of “The Baptism of Christ,” challenging the idea that his son and apprentices finished it.
“The Baptism of Christ” has always felt like a scene in motion—John the Baptist pouring water forward while Jesus glows as heaven opens behind them.
Now, an AI-assisted investigation is testing a question that has quietly haunted connoisseurs and historians for years: how much of the work should be attributed to El Greco himself, and how much to others in his workshop.
AI at the paint-bristle scale
The study centers on microtexture—tiny physical patterns in the paint surface that can reflect how an artwork was made.. Instead of relying primarily on visible brushwork or broader stylistic impressions. the researchers examined “The Baptism of Christ” at a resolution fine enough to capture features linked to individual brush bristles.
The goal wasn’t just to label the painting as “El Greco” or “workshop.” It was to ask whether the painting’s material behavior is consistent across its parts—an approach that can, in theory, reveal whether multiple hands were truly working in a way that leaves measurable differences.
In the Renaissance, master painters often worked with apprentices.. Helpers might mix pigments, prepare surfaces, and complete portions of scenes while learning under a studio’s established methods.. Yet for masterpieces whose paperwork is missing—as is the case for many workshop-era works—pinning down exactly who painted what can be a puzzle with high stakes: attribution can reshape an artwork’s historical meaning and. practically. its market and cultural value.
What the model learned—and what it challenged
To train the machine-learning model, the researchers used a dataset built from 25 paintings by nine student artists. With that training, the system learned textural “signatures” associated with brushwork at the microscopic level.
When the AI then analyzed two paintings connected to El Greco—“Christ on the Cross with Landscape” (generally thought to be solely El Greco) and “The Baptism of Christ” (the one tangled in multipainter debate)—the results separated in an important way.. The model’s interpretation of “Christ on the Cross with Landscape” aligned with expectations: it detected a more uniform authorship signal.
But “The Baptism of Christ” produced a different kind of pattern.. Rather than strongly splitting the surface into distinct communities of different painters. the analysis suggested a deeper continuity across segments that art historians previously treated as potentially finished by someone else.. In other words, the microtexture appeared more interconnected—and more uniform—than the multipainter hypothesis would predict.
The researchers describe this as “muddied waters” rather than a final verdict. Even if the painting shows less evidence of clear hand-to-hand changes than expected, aging, changes in tools, or other physical factors can still blur the boundary between “one artist” and “multiple contributors.”
Why experts are cautious
Independent experts reviewing the approach stressed that AI doesn’t operate like a magic eye that always sees truth on first attempt.. Artists can alter their styles over time, and collaborators might consciously imitate a master.. Conservation treatment, surface damage, and how materials age can also influence measurements the model reads.
Critically, the validation base matters.. If the training and testing pool is limited—especially if the model is trained on student paintings rather than broader historical material—there’s a risk that the system performs well on cases that resemble its training. while remaining uncertain on older works with different production conditions.
One concern raised by outside researchers is that the evaluation should be tested more extensively on real, unseen paintings where the attribution ground truth is already known. Without that wider validation, even thoughtful models can give confident interpretations that are hard to fully trust.
The human story behind the science
For art history, this debate isn’t just academic taxonomy. A conclusion about authorship changes how historians interpret El Greco’s final years—what he intended to finish, how his workshop operated, and how artistic labor was distributed in a studio at the time of his death.
If future work supports the new microtexture findings. it could mean that “The Baptism of Christ” reflects more of El Greco’s own hand than previously believed.. That might shift attention toward how he worked late in life. including the possibility that he used different brushes or that his hands were influenced by the physical realities of aging.
Even if the final answer stays complex, the bigger impact is methodological: the study shows how microscopy combined with machine learning can add a new layer to the attribution toolkit—one that can complement, not replace, careful visual scholarship.
Where AI could go next in attribution
The researchers behind the model say their tool isn’t meant to replace art historians. Instead, it’s positioned as a way to help narrow questions and guide further examination—especially when attribution disputes hinge on faint physical clues that the naked eye can’t reliably settle.
Looking ahead. the same approach could be used to compare other works linked to El Greco’s studio. potentially searching for consistent microtexture patterns that might correspond to lesser-known individuals in workshop networks.. That kind of “signature hunting” is compelling precisely because workshop art often leaves behind names in records that don’t always match the physical evidence on the surface.
For now. the safest interpretation is also the most honest: this research doesn’t end the debate. but it changes its shape.. In the long-running mystery of “The Baptism of Christ. ” AI has helped reopen the conversation—by zooming in so far that even old assumptions may need to be repainted. one bristle at a time.
8-year-old’s “Rise” became Artemis II zero-gravity indicator
Sharks in the Bahamas tested positive for drugs—what it means