What is it about?
For more than two hundred years, one of Hans Holbein’s drawings in the Royal Collection has been identified as Anne Boleyn, largely because of an eighteenth-century inscription. This study asks whether that traditional identification is reliable. The research combines art history, documentary evidence, conservation evidence, and facial recognition technology. Instead of relying only on later labels, the study compares Holbein’s life-drawn sketches with historical descriptions of Anne Boleyn and with known family relationships in the Tudor court. The authors use facial recognition to measure patterns of facial similarity between drawings and portraits, including the authenticated portrait of Anne’s daughter, Elizabeth I. The results suggest that the drawing traditionally labelled as Anne Boleyn may not be the strongest candidate. Another Holbein drawing, currently catalogued as an unidentified woman, better matches contemporary descriptions of Anne as dark-haired and slender. It also shows a strong facial similarity pattern with Elizabeth I and with the wider Boleyn–Howard family network. The study does not claim that facial recognition can prove identity on its own. Instead, it shows how computational analysis can provide useful additional evidence when combined carefully with historical research.
Featured Image
Photo by McGill Library on Unsplash
Why is it important?
This work is important because Anne Boleyn remains one of the most recognisable and contested figures in English history, yet no completely secure lifetime painted portrait of her is known. Much of what the public recognises as Anne’s image may depend on later portraits, copies, inscriptions, and inherited assumptions. The study is timely because museums and scholars are increasingly using digital tools to revisit long-standing questions in cultural heritage. This research shows how facial recognition can be applied responsibly to historical portraiture by focusing on Holbein’s working sketches, which were made as practical likenesses rather than symbolic or idealised images. What is distinctive about this work is its interdisciplinary approach. It does not treat AI as a replacement for art historians. Instead, it combines contemporary written evidence, provenance, material features, Holbein’s working methods, and biometric comparison. This provides a more transparent way to test competing attribution hypotheses. The research could influence how museums, historians, and the wider public understand Anne Boleyn’s visual legacy. It also offers a model for reassessing other uncertain historical portraits where traditional evidence is fragmentary, contradictory, or based on later labelling.
Perspectives
This study shows that computational methods can add real value to art-historical research when they are used with caution and historical discipline. Facial recognition is most useful here because it measures underlying facial structure rather than features such as hairstyle, costume, or later artistic convention. At the same time, the paper makes clear that numerical similarity scores should not be treated as final proof. Historical portraiture is affected by pose, age, artistic style, medium, condition, and copying history. For that reason, facial recognition must be interpreted alongside documentary evidence, material analysis, and expert judgement. The broader perspective is that AI can help reopen important questions in cultural heritage, especially where accepted identifications rest on weak or unverified evidence. In this case, the findings suggest that a drawing long accepted as Anne Boleyn may need reassessment, while another unidentified Holbein drawing may deserve serious consideration as a possible likeness of Anne herself.
Professor Hassan Ugail
University of Bradford
Read the Original
This page is a summary of: Reassessing Anne Boleyn and other Boleyn women in Holbein drawings using facial recognition, npj Heritage Science, March 2026, Springer Science + Business Media,
DOI: 10.1038/s40494-026-02456-0.
You can read the full text:
Contributors
The following have contributed to this page
