Saturday 21 November 2026SearchSign in
Ancient Herald

News from the ancient world

Topics

AI in archaeology

Artificial intelligence in archaeology mostly means machine learning: computer models that learn to recognise patterns from thousands of labelled examples. Archaeologists use them to search satellite images for sites, to read damaged or hidden texts and to sort finds such as pottery, seeds and stone tools. A person still has to check what the model proposes, in the field or in the archive.

Our stories show both the reach and the limits. In Sardinia, a model trained on 1,612 satellite images of known nuraghes led to six probable new towers, along with eight false alarms. A University of Bologna model retrained on 1960s CORONA spy photos found four new sites west of Baghdad that fieldwork then confirmed. X-ray scans and AI recovered nearly 1.5 metres of text from PHerc. 1667, a charred scroll from Herculaneum, without unrolling it. Some tools are still unproven: Apollo Restore, which fills gaps in Greek texts, is not yet peer reviewed.

18 stories, newest first.