Background: Orthodontic diagnosis and treatment planning have traditionally relied on manual cephalometric analysis, which is time-intensive and subject to operator variability. Artificial intelligence (AI) offers promising avenues for automation, greater accuracy, and reproducibility across clinical workflows.
Objective: To systematically evaluate current AI-based frameworks applied to orthodontic diagnosis, cephalometric landmark detection, malocclusion classification, and treatment outcome simulation.
Methods: A structured review and comparative study design was employed, integrating data from peer-reviewed studies (2016–2024) and evaluating AI model performance using metrics including diagnostic accuracy, prediction precision, and treatment efficiency.
Results: Deep learning models—particularly convolutional neural networks (CNNs), Vision Transformers, and hybrid architectures—achieved cephalometric accuracy rates of 88.6–95.2%, reduced landmark identification time by up to 77.7%, and improved diagnostic consistency over traditional methods.
Conclusion: AI-integrated orthodontic systems demonstrate strong clinical potential. Standardisation of datasets, validation across diverse populations, and regulatory frameworks remain prerequisites for widespread implementation.