The results are results, isn't it?
Five centuries after Anne Boleyn's execution, her face remains one of history's most contested absences — a void that has now drawn computer scientists into a discipline long guarded by art historians. A University of Bradford team applied machine-learning facial recognition to Hans Holbein's Tudor drawings, believing they have identified a mislabeled sketch of the queen through patterns of family resemblance. The art history establishment has responded with sharp skepticism, and the dispute illuminates something older than any algorithm: the enduring human struggle over who gets to say what the past looked like, and by what means.
- Anne Boleyn left behind no confirmed lifetime portrait, making her face one of history's most tantalizing unsolved puzzles — and a magnet for competing claims of discovery.
- A computer science team, including a researcher who funds her historical passion by working as a cleaner, used facial recognition to flag an 'unidentified woman' in the Holbein collection as Boleyn — a finding that shocked even its own authors.
- Art historians are pushing back hard, with one calling the research 'a load of rubbish,' arguing that algorithms built for modern photographs cannot be trusted against 500-year-old drawings.
- The National Portrait Gallery undercuts the study's foundation entirely, noting there is no securely authenticated reference portrait of Boleyn to begin with — making any algorithmic comparison inherently unstable.
- The Royal Collection Trust, which holds the Holbein drawings, has quietly distanced itself from the findings, while the researchers stand firm behind their peer-reviewed methodology.
- The clash is landing not as resolution but as a widening fault line between computational and traditional ways of knowing history — with Boleyn's face still, stubbornly, unresolved.
Anne Boleyn lost her head in 1536, and five centuries later no one can agree on what she looked like. Every portrait we have was made after her death. The sketches that might have captured her living face are either uncertainly labeled or buried somewhere in Hans Holbein the Younger's vast collection of Tudor drawings. Now a University of Bradford computer science team believes it has solved the riddle — and the art history world is furious.
Led by Prof Hassan Ugail, the team ran digital copies of the Holbein drawings through a machine-learning algorithm that compared facial features across the collection, looking for family resemblance patterns linking potential Boleyn portraits to sketches of her cousins and her daughter Elizabeth. The algorithm identified clustering around a sketch long labeled simply as an 'unidentified woman.' Lead author Karen Davies, who has been working as a cleaner since 2024 to fund her historical research, had long doubted the authenticity of the sketch most art historians accept as Boleyn — its inscription written in an 18th-century hand, its subject fair-haired and full-chinned, at odds with historical descriptions. A chance conversation with a client eventually connected her to Ugail. 'If evidence can be tested, then it should be tested,' she says.
The art history establishment is unconvinced. Dr Bendor Grosvenor dismisses the research bluntly, arguing that facial recognition technology designed for modern photographs cannot be reliably applied to 500-year-old drawings, and questioning whether peer review adequately scrutinized the methodology. He remains confident the traditionally accepted Holbein sketch is genuine, suggesting the blonde hair may be a topcoat worn away over time and the informal dress a marker of high status — the sketch perhaps capturing Boleyn during pregnancy.
Dr Charlotte Bolland of the National Portrait Gallery identifies a more fundamental problem: there is no securely authenticated lifetime portrait of Boleyn to serve as a reference point at all. Her reign lasted only three years before her execution on charges of adultery, incest, and treason — too brief, perhaps, for a stable image to take hold. Some portraits may have been deliberately destroyed after her fall. The Royal Collection Trust, which holds the Holbein drawings, did not participate in the study and does not endorse its findings.
What the dispute ultimately reveals is a collision between two ways of knowing the past. The computer scientists argue that an algorithm, free from human bias, can detect patterns invisible to the trained eye. The art historians counter that art cannot be reduced to facial geometry, that context and scholarly tradition carry weight no machine can properly measure. Bolland puts it more gently: curiosity drives people to keep throwing new methodologies at questions that have resisted answers for centuries. Whether a computer can finally reveal Boleyn's face remains, for now, a question without consensus.
Anne Boleyn lost her head in 1536, but five centuries later, she remains a figure of obsession—not least because no one can quite agree on what she looked like. Every portrait we have of her was painted after she died. The sketches that might have captured her living face are either labeled with uncertainty or hidden somewhere in the vast collection of Tudor drawings created by Hans Holbein the Younger, the masterful artist who moved through Henry VIII's court. Now a team of computer scientists believes they have solved the riddle using facial recognition software, and the art history world is furious.
The University of Bradford research team, led by Prof Hassan Ugail, took digital copies of all the Holbein drawings and ran them through a machine-learning algorithm designed to compare facial features across the collection. The system looked for family resemblance patterns—comparing potential portraits of Boleyn to sketches of her first cousins and her daughter Elizabeth, the future queen. When the algorithm found clustering, a mathematical measure of facial similarity, the researchers interpreted it as evidence that they had located a sketch of Boleyn that had been sitting in the collection for centuries, mislabeled as an "unidentified woman." The result, according to Karen Davies, the lead author, "shocked us completely."
Davies came to this work through an unconventional path. Since August 2024, she has been working as a cleaner to fund her true passion: historical research. She had long doubted that the Holbein sketch actually labeled "Anne Boleyn"—the one most art historians accept as genuine—was authentic at all. The label was written in an 18th-century hand, centuries after Boleyn's death. The woman in the sketch wore informal dress, had light-colored hair, and possessed a full chin, all details that seemed to contradict historical references to her dark hair and thin neck. When Davies mentioned her passion project to a client, that conversation led her to Ugail. "If evidence can be tested, then it should be tested," she says.
But the art history establishment is not convinced. Dr Bendor Grosvenor, a respected art historian, calls the research "a load of rubbish" and "a load of old phooey." He argues that facial recognition technology, designed for modern photographs, cannot be reliably applied to 500-year-old drawings. He questions whether the peer-review process adequately scrutinized the methodology and the researchers' interpretation of the algorithm's output. He remains convinced that the traditionally labeled Holbein sketch is genuine—that it was identified by someone who knew Boleyn personally, even if the visible inscription came later. The informal dress, he suggests, was worn only by those of high status. The blonde hair may have been a topcoat that rubbed away over time. The sketch, he believes, captures Boleyn during pregnancy, an intimate moment.
Dr Charlotte Bolland, a senior curator at the National Portrait Gallery, points to a deeper problem: there is no agreed-upon reference portrait of Boleyn to begin with. "We don't have a lifetime painted portrait of her that's absolutely secure, a wonderful painting that we can use as a reference point," she says. Boleyn's reign lasted only three years before her execution on charges of adultery, incest, and treason. That brief time may not have been long enough for a stable, widely recognized image of her to take hold. There is also the tantalizing possibility that some of her portraits were deliberately destroyed after her fall from grace.
The Royal Collection Trust, which holds the Holbein drawings, did not participate in the study and does not endorse its findings, though it welcomes scholarly engagement with its collection. Ugail remains unmoved by the criticism. "The results are results, isn't it?" he says. The research has passed peer review, a rigorous gatekeeping process that lends it academic credibility. Yet Grosvenor suspects that many of his art history colleagues are too nervous to speak out publicly against the findings.
What the debate reveals is a fundamental tension between two ways of knowing the past. The computer scientists argue that an algorithm, free from human bias and existing labels, can detect patterns invisible to the eye. The art historians counter that works of art cannot be reduced to facial geometry, that context matters, that the written record and scholarly tradition carry weight that no machine can properly weigh. Bolland frames it differently: "There is this curiosity and I think that drives ongoing research—people throwing different methodologies to try and answer a question that has been thought about for hundreds of years." The emotional tragedy of Boleyn's life, the mystery of her true appearance, continues to pull researchers toward her. Whether a computer can finally reveal her face remains, for now, a question without consensus.
Citações Notáveis
We don't have a lifetime painted portrait of her that's absolutely secure, a wonderful painting that we can use as a reference point.— Dr Charlotte Bolland, National Portrait Gallery
If evidence can be tested, then it should be tested.— Karen Davies, lead researcher
A Conversa do Hearth Outra perspectiva sobre a história
Why does it matter so much what Anne Boleyn looked like? She's been dead for nearly five hundred years.
Because we have almost nothing authentic from her lifetime. Every painted portrait came after she died. That absence creates a kind of vacuum that people keep trying to fill—with sketches, with speculation, with whatever evidence they can find. It's not really about vanity. It's about the possibility of connection across time.
And the computer scientists think they've solved it by comparing faces?
They think they've found a pattern. They ran the algorithm against portraits of her relatives—her cousins, her daughter Elizabeth—and found mathematical clustering, a measure of facial similarity. If the faces cluster, they argue, it suggests family resemblance. That's how they claim to have identified the hidden sketch.
That sounds reasonable. Why are the art historians so angry?
Because they're asking a different question. They want to know whether a machine trained on modern photographs can actually read a 500-year-old drawing. Whether facial geometry alone can tell you who someone is when you have no secure reference point to begin with. And whether five centuries of scholarship should be overturned by an algorithm.
Is there a secure reference point?
That's the problem. No. There's a sketch labeled "Anne Boleyn" that most historians accept, but even that's contested. The label was written long after her death. The details don't quite match what we know from written records. So you're trying to identify someone using a method that requires you to already know what they look like.
A catch-22.
Exactly. And the art historians point out that the researchers are comparing drawings to drawings, not photographs to photographs. A sketch is an interpretation. It's not a neutral record. The artist made choices about what to include, what to emphasize, what to leave out.
So who's right?
That's what makes this interesting. The research passed peer review. It's rigorous in its own way. But it's asking a question that art history has been asking for centuries, and it's using a tool that art history isn't sure it trusts. Both sides have legitimate points. The real question is whether we're ready to let machines reshape how we think about the past.