@article{JMAI11282,
author = {Anmar Arab and Abubaker Qutieshat},
title = {Artificial intelligence in orthodontics: a narrative review of clinical and educational applications},
journal = {Journal of Medical Artificial Intelligence},
volume = {9},
number = {0},
year = {2026},
keywords = {},
abstract = {Background and Objective: Although artificial intelligence (AI) in orthodontics has been addressed in several recent reviews, much of the literature has focused on selected clinical applications, technical principles, or ethical considerations rather than providing a single integrative synthesis across clinical practice, workflow, education, and advanced modalities. This review synthesises contemporary evidence on how AI is being used in orthodontic practice and training and maps its current capabilities, benefits, and limitations across fourteen application domains grouped into four clinician-relevant categories.Methods: An integrative narrative review approach was adopted. Scopus was searched for English-language publications from 2020 to 2025 using title-based combinations of orthodontic terms with artificial intelligence, machine learning, and deep learning. Retrieved records were screened for substantive relevance to orthodontic AI applications, and eligible studies were coded and thematically mapped into fourteen domains grouped under diagnosis and assessment, treatment planning and prediction, clinical workflow and monitoring, and education and advanced modalities.Key Content and Findings: AI demonstrates strong performance in automated cephalometric landmarking and classification tasks, with mean radial errors (MREs) commonly around 1.1–1.8 mm, success detection rates (SDRs) often 70–87% within 2 mm and above 90% at 4 mm, and malocclusion-classification accuracies frequently in the 90–97% range. In treatment planning, AI contributes to extraction and orthognathic decision support, virtual setup optimisation, and predictive modelling of tooth movement, facial soft-tissue response, and airway implications. In clinical practice, AI-enabled remote monitoring supports earlier detection of tracking failures and appliance-related issues, with recent studies reporting 93.2% sensitivity and 86.2% specificity for aligner seat-versus-unseat detection, alongside high diagnostic performance for common fixed-appliance emergencies. Emerging educational applications include AI-assisted simulation, adaptive learning, and competency assessment, often integrated with virtual reality (VR) and augmented reality (AR) environments.Conclusions: Current evidence indicates that AI can enhance orthodontic efficiency, consistency, and personalisation, particularly when embedded within digital workflows. However, generalisability, external validation, data bias, and governance remain critical barriers, reinforcing the need for robust validation, diverse datasets, and careful integration that preserves clinical judgement.},
issn = {2617-2496}, url = {https://jmai.amegroups.org/article/view/11282}
}