<![CDATA[Background: Histopathological (HPA) analysis is the gold standard for oral lesions. However, clinical diagnoses often have low agreement rates (44.1%). Artificial intelligence (AI) may offer diagnostic assistance, but its accuracy requires critical evaluation. Purpose: This study compared the diagnostic accuracy of AI models, ChatGPT and Gemini, in interpreting HPA images of oral maxillofacial lesions to the gold standard diagnosis from Oral and Maxillofacial Pathology (OMP) specialists. Methods: This analytical observational study used 54 digital HPA slides from a university research center. The diagnoses generated by ChatGPT and Gemini were evaluated for agreement ('Correct' or 'Incorrect') with the definitive diagnoses made by OMP specialists. Ethical approval was obtained. Results: Gemini demonstrated a diagnostic accuracy of 74.07% (40 of 54 cases), while ChatGPT achieved 70.37% (38 of 54 cases). The most common lesions were mucoceles and dentigerous cysts. A statistically significant difference (<0.001) was observed between the accuracy of both AI models and the OMP specialist. Conclusion: AI models showed considerable ability to recognize histopathological patterns, but their accuracy was significantly lower than OMP specialists. AI is an augmentative tool for triage or learning but cannot replace the role of OMP specialists in establishing a definitive diagnosis.]]>
Copyrights © 2026
