For nearly five centuries, the true appearance of Anne Boleyn has remained one of Tudor history’s enduring mysteries.

Now, an ambitious facial recognition study has reopened that mystery by examining drawings created inside Henry VIII’s court.

Researchers believe an unidentified woman portrayed by Hans Holbein the Younger could potentially be England’s most controversial queen.

Their findings also suggest surprising identities for other portraits traditionally associated with the extraordinary women of the Boleyn family.

Yet behind the attention-grabbing possibility lies a complicated debate about technology, historical evidence, and the reliability of Renaissance portraiture.

The central question is deceptively simple: Can a computer identify someone whose authenticated facial appearance remains largely unknown?

For historians studying Anne Boleyn, answering that question requires confronting centuries of missing records and disputed artistic identifications.

The new investigation offers intriguing possibilities, but its conclusions remain provisional, and several important limitations deserve careful examination.

Anne Boleyn became Henry VIII’s second wife in 1533, following years of political turmoil surrounding his first marriage.

Her relationship with the king transformed England’s religious landscape and helped reshape the monarchy’s relationship with papal authority.

However, the queen’s extraordinary rise ended abruptly when she was accused of adultery, incest, and plotting against Henry.

She was executed at the Tower of London on May 19, 1536, following proceedings widely questioned by historians.

Although her reputation survived through historical records and later portraits, identifying an unquestionably authentic contemporary likeness remains difficult.

That uncertainty creates an unusual problem because Anne remains among the most recognizable women in British royal history.

Millions recognize the familiar dark-haired woman wearing a French hood, pearl necklace, and distinctive letter B pendant today.

Yet the surviving paintings associated with this appearance were generally created decades after the queen’s execution in 1536.

The separation between Anne’s lifetime and her familiar painted representations has encouraged generations of speculation about her face.

Some researchers believe certain surviving images preserve lost contemporary originals, while others question how faithfully those images represent her.

The recently published facial recognition study attempts to investigate that uncertainty using computational methods alongside traditional historical research.

Its authors are Karen L. Davies, Hassan Ugail, and David G. Stork, whose backgrounds encompass historical investigation and technology.

Their research appeared in March 2026 in the peer-reviewed journal npj Heritage Science, attracting wider attention beyond academic circles.

Rather than announcing definitive identification, the researchers proposed reconsidering a previously unidentified drawing as a possible likeness of Anne.

That distinction matters because the difference between a promising hypothesis and a proven historical discovery remains enormous here.

The drawing at the center of the investigation belongs to the Royal Collection and carries identification number RCIN 912190.

It portrays a dark-haired woman wearing expensive clothing and an unusual headdress, looking directly toward the viewer.

Her facial expression appears restrained, while her carefully depicted clothing suggests someone familiar with an affluent social environment.

The sitter possesses dark eyes, arched eyebrows, and a relatively narrow build, characteristics broadly consistent with descriptions of Anne.

Holbein also included annotations describing luxurious fabrics, providing additional information about the clothing the unidentified woman may have worn.

These details establish that the sitter could have belonged to the prosperous social circles surrounding Henry VIII’s court.

However, expensive clothing alone cannot distinguish a queen from the numerous noblewomen and wealthy courtiers living in Tudor England.

Holbein worked among influential families who commissioned portraits for reasons extending far beyond preserving the appearance of royalty.

A beautifully dressed woman in his surviving drawings might therefore represent any number of people from those circles.

The drawing’s possible date provides another reason researchers have considered connecting the mysterious sitter to Anne Boleyn.

Holbein returned to England in 1532 and continued producing remarkable portraits until his death in 1543.

Anne occupied the English throne between 1533 and 1536, placing her within the chronological boundaries of that artistic period.

Nevertheless, those dates encompass many other women whose identities could fit the drawing’s broad physical and social characteristics.

Dating an artwork to Anne’s lifetime establishes possibility, but it does not independently establish who actually sat for Holbein.

The researchers therefore turned toward facial recognition software, hoping to identify patterns not immediately apparent through conventional visual comparison.

Their approach examined eighty-one Holbein sketches alongside two additional portraits associated with Anne Boleyn and Elizabeth I.

These images provided a collection of faces, including known relatives and individuals without recognized close family connections.

Among the most important comparisons was a portrait showing Anne’s daughter, the future Elizabeth I, during adolescence.

That painting, traditionally attributed to William Scrots, depicts Elizabeth around twelve or thirteen years old in the 1540s.

The researchers investigated whether facial similarities might connect Elizabeth with the unidentified woman in Holbein’s surviving drawing collection.

Their software produced a similarity score of 76.9 percent between the brunette sitter and young Elizabeth’s portrait.

This result became a central feature of the argument supporting the unidentified woman’s possible connection to Anne Boleyn.

But a similarity score should never automatically be understood as the probability that two individuals share a relationship.

It describes a mathematical comparison within the software’s framework, requiring interpretation against controls, thresholds, and relevant validation data.

Without understanding that distinction, a percentage can appear considerably more decisive than the underlying historical evidence actually permits.

The researchers argued that their results were consistent with familial relationships when considered alongside the available documentary evidence.

They also emphasized that their findings should complement established historical methods rather than replace them with technological certainty.

One especially important limitation concerns the material being analyzed, because sixteenth-century drawings differ fundamentally from modern photographs.

Facial recognition systems often operate on photographic images that capture actual physical features through predictable optical processes.

Renaissance drawings instead record an artist’s interpretation, influenced by technique, available time, lighting, and the sitter’s preferences.

Even exceptionally realistic images may contain deliberate adjustments that alter the geometric proportions computational software attempts to measure.

Hans Holbein was an exceptionally skilled portraitist, celebrated for capturing subtle expressions and remarkably convincing individual facial characteristics.

His surviving drawings reveal careful observation, extraordinary technical control, and a sophisticated understanding of human appearance and personality.

However, exceptional artistic ability does not guarantee that every drawing reproduces its sitter’s facial measurements with photographic precision.

Holbein sometimes revised portraits after additional sittings, demonstrating that his working process could involve substantial changes and adjustments.

Those revisions might reflect improved observation, requests from patrons, or artistic decisions that cannot now be independently reconstructed.

Researchers favoring his preparatory sketches argue that such drawings were intended to preserve accurate information for finished portraits.

That is a reasonable argument, especially because artists depended upon preparatory images when creating more elaborate painted compositions.

Yet working accuracy and exact biometric accuracy are different standards, particularly when evaluating faces through modern computational methods.

A subtle adjustment to someone’s chin, eyebrows, or nose could potentially influence the numerical comparison between two representations.

Differences in drawing angles introduce further complications because facial geometry changes significantly depending upon the sitter’s apparent orientation.

Lighting, unfinished details, paper texture, and artistic materials can also affect the information available to facial recognition systems.

These concerns become especially significant when the number of reliably identified historical comparison images remains relatively limited overall.

The study’s collection included eighty-three images, but many lacked contemporary documentation conclusively establishing the identities of their sitters.

Consequently, the computer was not working with a large archive of verified photographs representing documented members of one family.

Instead, researchers were evaluating artistic representations containing different levels of uncertainty about identity, physical accuracy, and familial relationships.

That challenge does not automatically invalidate computational analysis, but it complicates any effort to interpret similarity scores confidently.

One revealing comparison involved Henry Howard, Earl of Surrey, whose surviving Holbein drawings show him in different poses.

The software assigned approximately 84.1 percent similarity between two images understood to depict the same historical individual.

Meanwhile, a portrait associated with his sister Mary Howard reportedly achieved an even higher similarity against one Henry drawing.

The apparent contradiction illustrates why these numerical results require careful interpretation rather than simple assumptions about biological resemblance.

A brother and sister can resemble one another closely, while portraits of the same person can differ artistically.

Nevertheless, such findings raise reasonable questions about the stability and comparative meaning of the resulting scores across images.

The unidentified brunette achieved an approximately 80.6 percent similarity score against a drawing identified as Mary Howard.

That result is especially interesting because Mary Howard belonged to Anne Boleyn’s extended family through their Howard relatives.

She was also a prominent Tudor noblewoman whose appearance Holbein had opportunities to observe during his English career.

Historian Kate Heard has previously suggested that the unidentified brunette drawing might be another study involving Mary Howard.

This possibility offers an alternative explanation for the resemblance without requiring the mysterious woman to be Anne Boleyn.

If both drawings represented Mary, their facial similarities would reflect a shared sitter rather than distant biological kinship.

However, that proposed identification also remains unproven, illustrating the broader difficulty of interpreting unattributed Renaissance portraits with confidence.

Another major complication concerns the limited number of authentic contemporary representations available for comparison with the unknown brunette.

The surviving 1534 medal of Anne Boleyn offers an especially valuable connection to her actual lifetime as queen.

It was produced while she was alive and features a profile traditionally accepted as her contemporary likeness.

Unfortunately, significant damage has affected important facial details, particularly areas crucial for sophisticated automated recognition and comparison procedures.

The researchers therefore excluded the damaged medal from their computational analysis, citing its inadequate condition for reliable measurements.

That decision has a defensible technical basis, but it also removes the strongest surviving contemporary facial reference.

Without such an anchor, an unknown portrait cannot be directly compared against an unquestionably reliable image of Anne.

Later portraits offer additional possibilities, although their creation decades after her death introduces another source of uncertainty.

One familiar representation belongs to the National Portrait Gallery and depicts Anne wearing her characteristic dark French hood.

The study incorporated this painting, which was produced during Elizabeth I’s reign rather than Anne’s own lifetime.

A comparison between that portrait and young Elizabeth reportedly produced a similarity score of approximately 84.4 percent.

The researchers interpreted relationships among these images as supporting their broader hypothesis about the unidentified Holbein drawing’s sitter.

Yet a later portrait might repeat artistic conventions or incorporate characteristics reflecting political expectations rather than actual appearance.

There is also the Hever Castle portrait, another important later representation traditionally understood to depict Anne Boleyn.

Although related to the familiar portrait tradition, its facial features are not identical to those of other versions.

Including this image might have provided an additional test of how consistently the software recognized presumed representations of Anne.

The researchers’ restricted selection of portraits consequently remains an important issue when assessing the overall strength of their conclusions.

The depiction of Elizabeth presents similar questions because children and adults can exhibit markedly different facial proportions.

Elizabeth’s adolescent portrait was compared with a woman thought to be considerably older, introducing developmental differences into the analysis.

An adult Elizabeth might have offered additional information about inherited facial structure, although later paintings introduce their own problems.

Images created during her reign frequently emphasized idealized political symbolism, presenting the monarch as youthful and exceptionally powerful.

However, Elizabeth’s surviving portrait tradition also includes representations made relatively early in her adult life and reign.

Such works could potentially support sensitivity testing, although their varying artistic styles would complicate direct computational comparisons considerably.

This tension highlights a methodological problem that becomes unavoidable when applying modern technology to historical artistic evidence.

Expanding the dataset might improve coverage, but adding stylistically different paintings could simultaneously introduce additional measurement errors overall.

Researchers must therefore balance the benefits of more comparisons against the risks of less consistent artistic material.

The debate becomes even more intriguing when considering another Holbein drawing traditionally labeled as Anne Boleyn by later collectors.

This second drawing, cataloged as RCIN 912189, portrays a fair-haired woman with a comparatively fuller facial appearance.

Its blonde hair and physical proportions contrast with the most frequently cited historical descriptions of Anne’s coloring.

The identification is also connected with a later inscription rather than secure documentary evidence from Anne’s lifetime.

For these reasons, researchers questioned whether the traditionally labeled drawing actually represents the unfortunate Tudor queen at all.

That skepticism has historical support because later inscriptions on Renaissance drawings have sometimes produced misleading or demonstrably incorrect identifications.

Yet the software generated another surprising result when comparing the blonde sitter with the unidentified brunette woman.

Their facial similarity score reportedly reached 78.3 percent, suggesting a mathematical resemblance despite their very different appearances.

The blonde woman also registered approximately 75.4 percent similarity with young Elizabeth, close to the brunette’s corresponding score.

If interpreted solely through those numbers, both women might appear comparably plausible as relatives of Elizabeth I.

The researchers proposed an alternative explanation involving Anne’s mother, Elizabeth Howard, who became Countess of Wiltshire and Ormond.

They suggested that the blonde portrait might depict Elizabeth Howard rather than her historically better-known daughter Anne Boleyn.

Such an identification would create an intriguing connection between two enigmatic Holbein drawings and the Boleyn family.

However, the hypothesis faces a significant chronological difficulty because Elizabeth Howard was substantially older than Anne during the 1530s.

She had married Thomas Boleyn by the beginning of the sixteenth century and died in 1538.

Even using conservative chronological estimates, Elizabeth would have been in her forties when Holbein returned in 1532.

The blonde sitter’s youthful appearance therefore creates legitimate questions about whether this proposed identification fits the historical circumstances.

Age cannot always be estimated accurately from portraits, particularly when artists employed flattering techniques or incomplete details.

Nevertheless, any convincing identification should explain potential contradictions rather than relying primarily upon similarities generated by software alone.

Further complications arise from heraldic imagery associated with the reverse of the drawing showing the blonde woman.

The presence of Wyatt family heraldry introduces another possible connection that cannot simply establish the sitter’s personal identity.

It might indicate ownership, historical association, or another purpose, but the precise explanation remains uncertain and disputed today.

The investigation also ventured beyond Anne and her mother by considering a separate drawing traditionally identified as Lady Vaux.

The authors explored whether this woman might instead be Mary Boleyn, Anne’s elder sister and former royal mistress.

Their suggestion drew partly upon computational comparisons and perceived similarities involving later representations associated with Mary Boleyn.

Yet the traditional Lady Vaux identification has meaningful contextual support within Holbein’s surviving body of portrait drawings.

Holbein also portrayed Sir Thomas Vaux, the woman’s husband, making a corresponding portrait of his wife entirely plausible.

Art historians have additionally identified material connections between relevant drawings, including characteristics of the paper used by Holbein.

Such physical evidence may offer valuable dating and contextual information independently of uncertain algorithmic facial resemblance measurements.

Mary Boleyn’s personal circumstances create another obstacle to assuming that she sat for Holbein around 1535.

Her marriage to William Stafford had angered the royal court and contributed to her exclusion from Anne’s circle.

Although exclusion does not make a portrait sitting impossible, it weakens assumptions about her continued access to court.

A later painting traditionally identified with Mary cannot independently prove that an earlier Holbein drawing portrays her either.

Later artists could borrow clothing styles or existing compositions without accurately preserving the identity of an original sitter.

These uncertainties illustrate how one disputed identification can become dependent upon several other equally disputed historical assumptions.

If the blonde woman is incorrectly assigned to Elizabeth Howard, conclusions about related faces could shift substantially thereafter.

Similarly, if the Lady Vaux drawing truly represents Lady Vaux, attempts to reinterpret it as Mary become less persuasive.

The broader study therefore depends upon combining computational patterns with documentary evidence capable of supporting specific historical identifications.

Its strength lies partly in opening established assumptions to examination rather than accepting inherited portrait labels without question.

Its weakness, critics argue, lies in whether the proposed identifications adequately address competing explanations and inconvenient contextual details.

Importantly, the researchers themselves did not claim that facial recognition had conclusively established Anne Boleyn’s true appearance.

Their published conclusions framed the brunette drawing as a candidate deserving reconsideration, subject to future evidence and revision.

That measured conclusion is considerably narrower than sensational suggestions that modern technology has finally solved the entire mystery.

Scientific research frequently advances understanding by identifying possibilities, even when the available evidence cannot support definitive conclusions yet.

Historical portrait identification remains particularly challenging because the necessary documentary records may have disappeared hundreds of years ago.

Artists sometimes produced drawings without recording names, while later owners attached identifications based upon uncertain traditions or guesses.

Portraits could also be copied, altered, damaged, relabeled, or separated from their original supporting documents over successive generations.

Every interruption in that historical chain makes reliable identification more complicated for researchers attempting reconstruction centuries afterward.

Computational methods may eventually help resolve selected mysteries when sufficient verified comparative material becomes available for rigorous testing.

However, meaningful progress will require transparent methodology, broad independent testing, and careful consideration of alternative historical explanations.

Researchers also need to establish how similarity scores behave across artists, media, poses, and different ages of sitters.

Without that information, impressive percentages risk becoming persuasive visual rhetoric rather than reliable historical identification evidence for scholars.

The case raises another important distinction between recognizing a person’s face and recognizing a family resemblance pattern.

Relatives sometimes look strikingly similar, but biological relationships do not guarantee predictable facial correspondence across every individual or generation.

Children inherit complex combinations of genetic traits, making resemblance variable even among siblings sharing the same parents.

Distant cousins create additional uncertainty because their shared genetic inheritance may produce little obvious similarity in facial appearance.

Furthermore, unrelated individuals can resemble one another strongly, especially when portraits emphasize fashionable styles and conventional facial presentation.

These overlapping possibilities complicate attempts to convert resemblance measurements directly into specific genealogical conclusions about historical individuals.

Ultimately, a convincing identification would benefit from independent evidence such as contemporary inscriptions, documented commissions, or verified provenance.

Such discoveries might support the facial recognition findings or reveal that an entirely different woman posed for Holbein.

Until then, the brunette sitter remains unidentified, despite the compelling possibility of a connection to Anne Boleyn.

The investigation nevertheless offers a valuable reminder that new technologies can generate worthwhile questions about familiar historical objects.

It also demonstrates why traditional scholarship remains essential when interpreting information produced by increasingly sophisticated computational analytical systems.

Holbein’s remarkable portraits continue to fascinate because they offer unusually intimate glimpses of people living nearly five centuries ago.

The faces seem immediate and recognizable, yet their identities and precise appearances can remain frustratingly difficult to establish.

Anne Boleyn’s story intensifies this fascination because her life ended violently while her historical legacy continued expanding afterward.

Her daughter became one of England’s most celebrated monarchs, while Anne herself became an enduring subject of fascination.

Each proposed portrait therefore carries significance far beyond artistic attribution, touching questions of reputation, memory, and royal identity.

The mysterious brunette woman might indeed be Anne, but the currently available evidence cannot establish that conclusion securely.

She might instead be Mary Howard or another noblewoman whose name disappeared from the drawing’s surviving documentation.

The blonde sitter and Lady Vaux proposals likewise remain hypotheses requiring stronger independent historical and art-historical confirmation.

For now, the surviving 1534 medal continues to provide the clearest securely contemporary visual connection to Anne.

What the investigation has accomplished is to challenge assumptions and encourage renewed scrutiny of Tudor portrait collections worldwide.

Perhaps additional research will uncover forgotten documentation or develop more reliable ways of comparing sixteenth-century artistic representations.

Until such evidence emerges, Anne Boleyn’s exact face remains tantalizingly elusive, even in an age of artificial intelligence.

The greatest revelation may be that technology can sharpen historical questions without necessarily providing the answers we seek.

And that uncertainty keeps the search for the real face of Anne Boleyn as compelling as ever.