The Face That History Erased: AI Reignites a 490-Year Debate Over Anne Boleyn's True Likeness

Nearly 490 years after her execution at the Tower of London, Anne Boleyn's face remains one of history's most contested blanks. No universally accepted portrait painted during her lifetime survives [1]. After her death in 1536, Henry VIII ordered a systematic campaign to destroy images of his second wife, and those who possessed her likeness had ample reason to dispose of it quietly [2]. Into that void, a team of researchers from the University of Bradford and Stanford University has now inserted an algorithm — and the results have split the world of Tudor scholarship.

A study published in March 2026 in npj Heritage Science argues that a Holbein sketch long labelled "An Unidentified Woman" is, in fact, Anne Boleyn — and that the drawing traditionally identified as her actually depicts someone else, most likely her mother, Elizabeth Howard [3]. The claim rests on AI-driven facial recognition, contemporary written descriptions, and a new methodology the authors call a "Working Likeness" framework. But art historians have pushed back hard, with one prominent critic calling the conclusions "statistical noise" [4].

The dispute is more than an academic footnote. It touches on fundamental questions about how technology interacts with historical evidence, who gets to define the appearance of iconic figures, and whether a "definitive face" for Anne Boleyn is even achievable — or desirable.

The Two Sketches

At the center of the controversy are two drawings from Hans Holbein the Younger's collection of preparatory sketches, now held in the Royal Collection Trust at Windsor Castle. Holbein arrived at the Tudor court around 1526 and spent years capturing its members in chalk and ink. Roughly 85 of his preparatory drawings survive, but only about 30 have been securely identified through contemporary inscriptions or corroborating evidence [4]. The rest were labelled in the 18th century, and many of those attributions are now questioned by modern scholars.

Source: npj Heritage Science / University of Bradford

Data as of Mar 1, 2026CSV

The sketch traditionally identified as Anne Boleyn — catalogued as RCIN 912189 — was given that label based on an unverified 18th-century inscription. It depicts a fair-haired woman with a substantial build and a pronounced double chin [3]. The second sketch, RCIN 912190, currently labelled "An Unidentified Woman," shows a dark-haired, slender figure — proportions that match the documentary record far more closely [1].

The Bradford Study's Method

The research team — led by independent historian Karen L. Davies, with Professor Hassan Ugail of Bradford's Centre for Visual Computing and Professor David Stork of Stanford — used a deep learning model called AdaFace, paired with what the authors describe as a Sample and Computation Redistributed Face Detector framework [3]. They compared the two sketches against a dataset of validated Tudor family portraits, looking for biometric patterns of familial resemblance.

Their headline finding: RCIN 912190 produced a 76.9% similarity score when compared to authenticated portraits of Elizabeth I, Anne Boleyn's daughter — a figure the researchers say falls within the expected range for close biological relatives [1]. The traditionally labelled sketch, RCIN 912189, did not produce comparable results.

Davies has been clear that the study is meant to open debate, not close it. "I think now we've opened up the question. It's not like we're making a claim and that's the thing settled," she told the Smithsonian [4]. Ugail echoed that framing: "We're not replacing scholarly judgement. What we've built is a quantitative test that can support or challenge attribution hypotheses" [1].

Stork argued the Holbein drawings are well-suited to this kind of analysis precisely because they were working documents rather than finished paintings. "The Holbein drawings functioned as working likenesses, technical blueprints for painted portraits," he said. "That makes them uniquely suited to biometric analysis, which measures bone structure and proportion rather than hairstyle or costume" [4].

The Skeptics Respond

The pushback has been swift. Art historian Bendor Grosvenor dismissed the study's conclusions on social media, calling them products of "statistical noise" and questioning whether facial similarity measurements between works by different artists in different media can yield reliable results [4].

Charlotte Bolland, Senior Curator at the National Portrait Gallery, highlighted a more fundamental problem: "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" [4]. Without a confirmed baseline image of Anne Boleyn, any AI comparison is measuring the distance between unknowns.

This criticism cuts to the heart of the methodology. Facial recognition algorithms are trained on photographs of living people. Applying them to 500-year-old chalk drawings — where the artist's style, the angle of the sitter, and the degradation of the medium all introduce variables — raises questions about whether the tool is being used within its domain of validity.

The Forensic Reconstruction Problem

The Bradford study is distinct from forensic facial reconstruction — the technique of building a face from skeletal remains — but both approaches share a common ambition: to recover a lost appearance. And both face similar epistemological limits.

Forensic facial reconstruction has a mixed track record. Population-specific soft-tissue models achieve root mean square errors of 3.3 to 4.4 millimeters, and operator errors in manual reconstructions range from 0.9 to 1.9 millimeters at different facial landmarks [5]. The general shape and position of major facial features can be estimated with reasonable accuracy because they are largely determined by the skull. But details like ear morphology, mouth shape, skin texture, and the effects of aging require artistic interpretation — introducing subjectivity that no algorithm can fully eliminate [5].

Source: OpenAlex

Data as of Jan 1, 2026CSV

The field's most celebrated success is Richard III. When his remains were discovered beneath a Leicester car park in 2012, Professor Caroline Wilkinson of the University of Dundee built a facial reconstruction from a high-resolution 3D skull scan. DNA analysis later confirmed the identity of the remains and revealed that Richard likely had blue eyes and light brown hair — contradicting centuries of dark-haired, dark-eyed portrayals influenced by Shakespeare [6]. The reconstruction worked because researchers had the actual skull. For Anne Boleyn, no such resource exists in usable form.

The Bones in the Tower

Anne Boleyn was buried in the chancel of the Chapel of St. Peter ad Vincula within the Tower of London. During restoration work in 1876, remains believed to be hers were exhumed and examined by Dr. Frederick J. Mouat, a former Professor of Medicine and Medical Jurisprudence [7].

Mouat described the bones as belonging to "a female of between twenty-five and thirty years of age, of a delicate frame of body, and who had been of slender and perfect proportions." He noted that the vertebrae were "particularly small, especially one joint (the atlas), which was that next to the skull, bearing witness to the Queen's 'little neck'" [7]. He found no evidence of a sixth finger or any malformation — countering a persistent myth about Anne's supposed deformity.

The remains were sealed in thick lead coffers, placed in oak boxes with copper screws, and reinterred beneath the chapel floor in 1877 [7]. They have not been examined since. No DNA testing has ever been conducted. Multiple campaigns and public petitions have sought to reopen the examination using modern forensic techniques, but Historic Royal Palaces and ecclesiastical authorities have consistently declined, citing respect for the sanctity of the burial site [7].

Without access to the actual skeletal remains, a true forensic facial reconstruction of Anne Boleyn — of the kind performed on Richard III — remains impossible. The Bradford study's approach, working from portraits rather than bones, is an alternative path, but one that carries fundamentally different limitations.

The Source Bias Problem

Any attempt to reconstruct Anne Boleyn's appearance must contend with the fact that contemporary written descriptions were rarely disinterested observations. Most accounts were produced by people with political, religious, or personal stakes in how she was perceived [8].

The most frequently cited physical description comes from a Venetian ambassador writing in 1532: "Not one of the handsomest women in the world. She is of middling stature, with a swarthy complexion, long neck, wide mouth, bosom not much raised, and in fact has nothing but the King's great appetite, and her eyes, which are black and beautiful" [8]. This account is detailed but embedded in Venetian diplomatic correspondence, where characterizations of foreign queens served political purposes.

Hostile sources went further. The Abbot of Whitby called her "a common stewed whore" in 1530 [8]. Imperial Ambassador Eustace Chapuys, loyal to Catherine of Aragon and the Habsburg interest, described Anne as "accursed." Catholic polemicists later claimed she had the attributes of a witch — a large mole, and possibly a sixth finger [8]. These details appear in sources written years or decades after her death and are widely regarded by historians as propagandistic embellishments rather than factual reporting.

Supporters offered a mirror image. Thomas Cranmer, Archbishop of Canterbury and a Reformation ally, praised her character and faith [8]. The French ambassador, representing a court that had known Anne during her years in France, contradicted claims that she dominated the king, reporting instead that she was "controlled by" Henry [8].

The Bradford researchers attempted to account for this bias by cross-referencing multiple descriptions and weighting those written closest to Anne's lifetime. But as the researchers themselves acknowledge, separating observation from agenda in Tudor-era sources is an inherently imprecise exercise.

A Woman Remade in Every Century's Image

The fixation on Anne Boleyn's face is part of a longer pattern. For five centuries, her image has been continuously reshaped to serve the purposes of whoever is doing the looking [9].

During the reign of Mary I, Anne was cast as a scheming temptress — the woman who had displaced the rightful queen and split England from Rome. When her daughter Elizabeth I took the throne, Anne was rehabilitated as a Protestant champion, the mother of a divinely ordained monarch [9]. Some Elizabethan-era portraits of Anne even incorporated Elizabeth's own features, grafting the daughter's face onto the mother to reinforce dynastic legitimacy [9].

In the Romantic era, Anne became a tragic heroine — beautiful, doomed, and wronged. Victorian historians treated her as a cautionary tale about the dangers of royal power. In the 20th and 21st centuries, feminist scholars reclaimed her as an intellectually formidable woman who was destroyed for challenging patriarchal authority [10]. Film and television — from Anne of the Thousand Days (1969) to Natalie Dormer's portrayal in The Tudors (2007–2010) — have consistently cast conventionally attractive white actresses, projecting modern beauty standards backward across half a millennium [9].

Each generation, in other words, has built the Anne Boleyn it needed. A "definitive" face, whether produced by AI or by any other method, risks collapsing that complex history into a single image — presenting as settled science what is, at best, an educated hypothesis.

Who Funded This, and Does It Matter?

The Bradford study does not disclose major external funding sources. Davies, the lead author, is an independent historian — she has spoken publicly about working as a cleaner since August 2024 to fund her research [11]. The computing resources came from Bradford's Centre for Visual Computing and Intelligent Systems. Stork's early explorations were partially supported by the Aby Warburg Institute in London [1].

Davies is also a published novelist and maintains a website where she writes about Tudor history. She has a book on Anne Boleyn-related evidence [11]. None of this constitutes a disqualifying conflict of interest, but it does place the research within a broader ecosystem of Tudor popular history, where public interest in the Boleyn story generates consistent commercial demand for books, documentaries, and exhibitions.

The timing of the study's media rollout — coinciding with the 490th anniversary of Anne Boleyn's execution on May 19, 1536 — is likely not accidental. Anniversary hooks are standard practice in academic publicity, but they also serve to maximize attention in a crowded media landscape.

What a Face Cannot Tell Us

The deeper question may not be whether RCIN 912190 depicts Anne Boleyn. It may be why we want so badly to know.

The desire to fix a historical figure's appearance — to resolve the ambiguity, to see the "real" person — is understandable. But it can also flatten the complexity of a life that was, by any measure, extraordinary. Anne Boleyn was a political actor who helped reshape the religious and constitutional landscape of England. She was a patron of reformist scholars, a mother whose daughter became one of England's most consequential monarchs, and a woman whose trial and execution remain subjects of legal and historical debate.

A face tells us none of that. And a face reconstructed from contested portraits, filtered through an algorithm trained on modern photographs, and interpreted in light of politically motivated 16th-century descriptions tells us even less than we might hope.

The Bradford study has value — not because it has settled the question, but because it has forced a fresh examination of evidence that had calcified into received wisdom. Whether RCIN 912190 is Anne Boleyn or not, the study has demonstrated that at least some of the traditional attributions of Holbein's sketches rest on shaky ground. That is a contribution worth making, even if the definitive face of Anne Boleyn remains, as it has for nearly five centuries, just out of reach.