Artificial Intelligence Applications in Dentistry: A Systematic Review

Background: Artificial intelligence technologies are increasingly integrated into dental practice, offering potential improvements in diagnostic accuracy, treatment planning, and patient outcomes. However, the extent and quality of evidence supporting these applications remain unclear. Methodology: We conducted a systematic literature search using PubMed, Cochrane Library, Embase, and IEEE Xplore databases from January 2015 to December 2024. Search terms included combinations of “artificial intelligence,” “machine learning,” “deep learning,” “dentistry,” “diagnosis,” and “treatment planning.” Studies evaluating AI systems in clinical or laboratory settings with measurable outcomes were included. Data extraction followed PRISMA guidelines, and methodological quality was assessed using the QUADAS-2 tool. Results: Twenty-three studies met the inclusion criteria. Most focused on diagnostic accuracy (n = 21), with few addressing treatment planning (n = 1) or outcome prediction (n = 1). Reported accuracies ranged from 82–94% for caries detection, 85–92% for periodontal disease assessment, and 88–96% for oral lesion identification. Orthodontic applications achieved 95–98% accuracy in cephalometric landmark identification, while implant planning studies demonstrated up to 96% agreement with expert strategies. Despite promising technical performance, 79% of studies were retrospective and conducted in controlled research settings, with limited external or prospective clinical validation. Risk of bias was highest in patient selection due to frequent use of case–control designs and archived imaging datasets. Conclusions: AI shows significant promise for enhancing dental diagnostics and treatment planning. However, most applications require further clinical validation before routine implementation. The disconnect between laboratory performance and real-world clinical validation represents a critical gap that must be addressed. Current AI systems should be viewed as diagnostic aids rather than replacements for clinical judgment. Practitioners considering AI adoption should understand current limitations and evidence quality, particularly the lack of prospective clinical validation in diverse populations.