<p>Artificial intelligence (AI) is being increasingly used in head and neck cancer care, with potential applications in screening, diagnosis, prognostication, and treatment planning. However, the clinical value of these tools is still uncertain because many models have limited validation and uncertain reproducibility. This review examined English-language studies published between January 2000 and March 2025 that applied machine learning or deep learning to the management of head and neck cancer. A systematic search of PubMed, Embase, Scopus, Web of Science, and the Cochrane Library identified 1,782 unique records, of which 65 met the inclusion criteria. Where three or more comparable studies reported the same outcome, pooled quantitative analysis was performed. The findings suggest that AI has shown useful performance across several areas of ENT oncology. PET-based imaging models differentiated HPV-positive from HPV-negative oropharyngeal cancer with an area under the curve of 0.83 (95% confidence interval, 0.79 to 0.86). Deep learning applied to digital pathology identified tumour invasion fronts with a Dice score of 0.88. Other models achieved high performance in segmentation of organs at risk (Dice 0.87 to 0.93) and in real-time lesion classification during laryngeal endoscopy, with sensitivity of 0.92 and specificity of 0.88. Models combining multiple data inputs performed better than single-input systems in predicting disease-free survival, with a concordance index of 0.79. Despite these encouraging results, important limitations remain, including small and demographically skewed datasets, inconsistent performance across institutions, limited interpretability, and evolving regulatory requirements. These limitations must be addressed before AI can be incorporated into routine clinical practice in head and neck oncology.</p>