Artificial intelligence (AI) research in dentistry has grown significantly in recent years. There have been some promising results regarding its translation into clinical practice. This paper aims to analyze, summarize, and discuss the results of studies addressing specifically AI in orthodontics diagnosis and treatment published in the last 5 years. Orthodontics is characterized by an average treatment duration between 18 and 24 months, prognostic uncertainty, and influencing factors that are difficult to model. To this end, the review has excluded papers on automated cephalometric landmarking or classification if not conducted within diagnosis and treatment. Namely, the review analyzes application domains, common applications and their datasets, and subfields of AI that have been reported in the literature. The review also considers how well AI has performed in the reported applications. The results have shown that the leading application domain is diagnosis and treatment planning, followed by assessment of growth and development, treatment monitoring, and evaluation of treatment outcomes. The datasets included various data such as cone beam computed tomography scans, cephalometric radiographs, intraoral clinical images, and different clinical data from case records. Machine learning is the most researched subfield of AI in the given application domains. As for the performance results, several studies have reported excellent multiple evaluation metrics. However, the results of most studies lack multiple metrics or have varied depending on the algorithm or model type/parameters when several models have been compared. By all means, the generalizability of applications is questionable due to typically small, unicentric, and non-standardized datasets.