Next-Generation Paleopathology: Using Commercial AI in Bioarchaeological Diagnosis

Artificial intelligence encompasses computational systems capable of performing cognitive functions such as learning, reasoning, and problem-solving. Within this domain, generative AI and large language models such as ChatGPT, Gemini, and Copilot have shown significant potential in clinical diagnostics. Their capacity to generate structured differential diagnoses, integrate radiological criteria, and synthesize complex information underscores their transformative potential. Despite extensive validation in modern medical disciplines, their application in paleopathology, defined as the study of ancient disease through skeletal bones, remains unexplored. This study assessed the diagnostic performance of three commercial AI systems (ChatGPT 5.0 Plus, Gemini 2.5 Flash, and Copilot 365) in two archaeological cases presenting distinct osseous lesions. Diagnostic outcomes were compared with interpretations by three expert anthropologists to evaluate concordance and reliability. Complete agreement was achieved for the diagnosis of a spinal lesion, whereas partial agreement was observed for a patellar lesion. The findings demonstrate that AI systems can generate plausible differential diagnoses but also highlight variability in interpretive reasoning and prompt sensitivity. Differences among platforms suggest distinct conceptual frameworks in diagnostic inference. Although current AI tools cannot replace expert assessment, their integration into paleopathology may enhance diagnostic reproducibility, encourage interdisciplinary dialogue, and support the development of domain-specific datasets to improve future model accuracy and reliability.