Abstract Artificial intelligence (AI) is increasingly applied in restorative dentistry, but its role in assessing dental composite restorations is not yet well established. This scoping review evaluated how AI has been used to assess mechanical, esthetic, and diagnostic aspects of composite restorations. A systematic search of PubMed, Embase, Scopus, and Web of Science from 2020 to July 30, 2025, identified studies that applied AI to assess dental composite restorations. Fourteen studies met the inclusion criteria, and all but one had a low to moderate risk of bias. Reported AI applications included tooth shade matching, evaluating mechanical properties, classifying restoration types from images and radiographs, predicting clinical performance, assessing cure depth and microtensile bond strength to estimate debonding risk, and identifying factors associated with marginal microleakage. The highest reported predictive accuracies were achieved by an artificial neural network predicting abrasive wear (99.7% accuracy), adaptive boosting and multilayer Perceptron models predicting flexural strength and Vickers hardness (up to 99.0 and 98.9% accuracy, respectively), and extreme gradient boosting predicting mechanical properties (98.8–99.6% accuracy), each on task- and dataset-specific internal validation. Overall, current AI models show promise for supporting composite restoration evaluation but remain at an early, proof-of-concept stage. The use of small, often single-center samples and the absence of external validation suggest that their performance metrics may be overly optimistic and not yet ready for clinical application. Larger, multi-center datasets, transparent reporting, external validation, and formal assessments of clinical utility are needed before these tools can be considered for routine clinical application.
Yılmaz et al. (Thu,) studied this question.