Recently, artificial intelligence (AI) capabilities in dentomaxillofacial radiology, especially in panoramic radiographs, have grown tremendously. The purpose of this review is to analyse and consolidate evidence published between 2020 and 2025 on the diagnostic accuracy of AI software for dental orthopantomographs. In its latest assessment, AI was shown to detect, number, and identify prosthetics and implants with sensitivities and specificities exceeding 90%. AI showed similar results in tooth detection, though some variability in performance is evident. Cavity recognition, as well as evaluating endodontic quality and identifying periapical lesions, tends to show the opposite performance pattern of sensitivity vs. specificity. It has been shown that AI performs best on well-contrasted, highly structured images, while subtle pathology detection and pediatric cases remain the most difficult. AI achieved over 85% accuracy in quantifying bone changes and stratifying systemic risk, demonstrating its capability for AI screening of bone loss and for evaluating osteoporosis risk. Although the results seem promising from the datasets controlled datasets, the issue of generalizability indicates the need for more extensive, heterogeneous datasets for training and external validation. In conclusion, AI is an adjunct to the clinician; however, the interpretation and diagnosis require significant support from trained professionals.
