AI Reveals Hidden Identity in Renaissance Master’s Tudor Sketches

Artificial intelligence has possibly solved a long-standing mystery surrounding two Renaissance sketches by the master portraitist Hans Holbein, suggesting that the identification of one portrait may have been incorrectly assigned some 300 years ago. Researchers at the University of Bradford have examined works from the Royal Collection and found that an unnamed woman depicted in one sketch could in fact be Anne Boleyn, Henry VIII’s ill-fated second wife, whilst the portrait long believed to show Boleyn may instead depict her mother, Elizabeth Howard. The finding, made using advanced AI technology that compares artistic features across Holbein’s extensive body of work, challenges assumptions that have persisted since the 1700s, when the sketches are believed to have been mislabelled.

The Enigma of Incorrect Attribution

The inconsistency concerning these two Holbein sketches originates chiefly in inconsistent labelling practices that affected the artist’s large collection of pieces. Researcher Karen Davies, who commenced the investigation, identified notable variations within the collection of more than 80 Holbein works contained within the Royal Collection. Her initial investigation, published in March, showed that fewer than 15 per cent of the sketches contain contemporary documentary verification—meaning most identifications depend on subsequent labelling rather than period documentation. This lack of rigorous historical documentation enabled incorrect attribution, particularly when labels were applied centuries after the works were executed.

The Windsor sketch in its own right displayed a puzzling contradiction to traditional attribution. The portrait shows a sitter with fair complexion and auburn locks, features that appeared at odds with documented accounts of Anne Boleyn, who was regularly recorded as having a darker complexion. Yet the sketch had been recognised as Boleyn’s portrait for generations, primarily since no one had methodically challenged the eighteenth-century attribution. Davies noted that the Holbein corpus was infamous for such errors—one famous case involved a portrait of Henry Howard that was incorrectly identified, when it depicted his father instead. These recurring mistakes indicated a broader problem with how the collection had been organised and maintained.

• Fewer than 15 per cent of Holbein works possess contemporary documentary verification.
• The Windsor sketch’s material properties contradicted conventional depictions of Anne Boleyn.
• 18th-century inscriptions created attributions that persisted for three centuries unchallenged.
• The Holbein corpus included recorded instances of mixed-up family member identities.

How Artificial Intelligence Uncovered the solution to an Art Historical Enigma

To address the labelling problem that had persisted for centuries, Davies partnered with Professor Hassan Ugail, head of the Centre for Visual Computing at the University of Bradford. Ugail had already showcased the capabilities of artificial intelligence in art history scholarship, having previously used his AI model to assign a painting to Raphael that had puzzled experts for decades. His knowledge of visual analysis and machine learning offered a valuable approach to reassess the Holbein corpus with fresh, objective scrutiny that human scholars alone might have missed.

The AI analysis proved remarkably effective at recognising patterns beyond the reach of traditional connoisseurship. By studying the entire collection of Holbein sketches and evaluating stylistic, compositional, and physiognomic features throughout the collection, the researchers constructed a comprehensive system of visual relationships. This data-driven process moved beyond subjective interpretation, instead uncovering clusters of works that possessed measurable similarities. The results challenged long-held assumptions and suggested that centuries of misattribution could ultimately be addressed through technological innovation.

The Matrix Strategy

Ugail described the technique behind their innovative analysis: the team systematically assessed each image against every other work in the Holbein corpus, creating an enormous visual database. The AI algorithm then identified which sketches possessed the most similar stylistic and visual affinities, organising similar works together. This algorithmic mapping uncovered unexpected groupings that contradicted conventional labelling, implying that the eighteenth-century inscriptions had fundamentally misorganised the collection.

The results were remarkable. The unidentified woman sketch clustered closely with other confirmed portraits of the Boleyn-Howard lineage, whilst the Windsor sketch seemed to display closer visual resemblance with portraits of Elizabeth Howard, Anne Boleyn’s mother. These findings pointed to a systematic reversal of identities, demonstrating that the wrong attribution had been made at a specific moment when the collection was rearranged and relabelled hundreds of years ago.

The Windsor Sketch and Its Complications

The Windsor sketch, one of two works central to this reassessment, has persistently posed art historians with a mystery that challenged conventional wisdom. For hundreds of years, it was firmly attributed as Anne Boleyn, the ill-fated second wife of Henry VIII, whose dramatic fall from grace and death in 1536 have fascinated scholars and the general public. Yet the visual features shown in the sketch appeared inconsistent with contemporary accounts of Boleyn herself. The sitter is shown in profile, displaying notably pale complexion and striking red hair—features that appeared inconsistent with historical accounts of Boleyn’s deeper skin tone and contrasting pigmentation.

Independent scholar Karen Davies began questioning this attribution whilst studying the wider Holbein corpus, which comprises more than 80 works spanning the Renaissance master’s prolific career. Her suspicions were strengthened by the discovery that the collection exhibited widespread attribution errors. A particularly glaring example concerned Henry Howard, the Earl of Surrey, whose portrait had been mistaken for that of his own father. These mistakes suggested that the 18th-century inscriptions, which had guided art historical understanding for generations, might be fundamentally unreliable and in requiring thorough reassessment.

Physical Features and Historical Documentation

Contemporary accounts of Anne Boleyn, drawn from courtly accounts and historical documents, consistently portrayed her as possessing a deeper skin tone than the figure shown in the Windsor sketch. Court observers recorded her notable characteristics and bearing, yet the auburn-haired, fair-complexioned woman depicted by Holbein seemed to depart from these recognised portrayals. This discrepancy concerned Davies and sparked more thorough examination into whether the sketch’s long-accepted attribution had been mistaken from the start or whether it had been wrongly labelled during the eighteenth-century reorganisation of the Royal Collection.

The wider scope of Holbein’s artistic practice and his careful attention to portraiture strengthens the case for the reassessment. The accomplished Renaissance painter was renowned for his careful observation and skilled technique, capturing likenesses with remarkable accuracy. If the Windsor sketch truly represents Elizabeth Howard, Boleyn’s mother, rather than Anne herself, it would represent a substantial revision to established scholarly consensus. Such a revelation would require thorough investigation of additional pieces in the corpus, conceivably opening up a cascade of formerly overlooked identities and relationships within the Tudor court.

Holbein’s Turbulent Period in the Tudor Court

Hans Holbein the Younger came to England at a time of particularly volatile periods, when Henry VIII’s marital upheavals and religious upheaval were transforming the nation’s political landscape. The German-born artist became the king’s favoured portraitist, documenting the Tudor dynasty with unprecedented closeness and privileged insight. Yet his standing in the royal household was far from secure; he operated within perpetual scheming, executing portraits of figures who might lose favour—or face the executioner’s block—within months. His sketches preserved far more than likenesses but fleeting moments of authority and fragility during a reign marked by sudden shifts in fate.

The sketches under examination were made throughout a era in which Holbein’s services were in greatest demand, as Henry VIII sought to immortalise his multiple wives and courtly companions. Working with remarkable speed and precision, the master generated a extensive collection of preparatory sketches and finished portraits that functioned as both artistic records and political records. His meticulous attention to detail and his capacity to convey the psychological essence of his sitters made his work invaluable to the Tudor establishment. Yet the considerable quantity of his work, combined with the turbulent conditions of the period, gave rise to confusion and false attribution in following centuries.

• Holbein recorded Henry VIII’s court amid considerable political instability
• His access to royal subjects was exceptional for a artist of foreign origin
• Many of his sitters faced execution or exile shortly after being portrayed
• He completed over eighty works in his extensive Tudor court collection
• His sketches served as both artistic records and tools of royal propaganda

Paving the Way for Research to Come

The consequences of this algorithmic assessment reach well beyond the two sketches in question. By demonstrating that digital techniques can effectively contest long-standing attributions and misidentifications, the research opens significant possibilities for re-evaluating the entire Holbein corpus and comparable holdings held in major institutions. Karen Davies has been cautious in presenting her findings rather than as final determinations but as invitations to scholarly debate, stressing that the analysis raises questions rather than resolving them definitively. This measured approach embodies best practice in art historical scholarship, where evidence needs to be thoroughly investigated and competing interpretations examined before consensus can be reached.

The Royal Collection Trust has acknowledged the enduring doubt regarding the unnamed sitter’s identification and endorsed the innovative approach that machine learning contributes to the question. As organisations globally contend with vast holdings of inadequately recorded heritage works, this project demonstrates how computational methods can function as a powerful tool for recovery and scholarly revision. Coming generations of experts may now utilise similar AI methodologies to further contested pieces, conceivably revealing connections and identities that have stayed hidden for decades. The collaboration of traditional experts and digital specialists points to a promising new frontier in Renaissance studies.