The treatment of laryngeal cancer has seen a shift towards organ preservation strategies with non-surgical treatment offering equivalent survival outcomes. Nonetheless, salvage total laryngectomy (SLT) remains an important curative management option in cases of treatment failure or recurrence. The authors applied machine learning (ML) models in the form of multiple classification algorithms trained to predict whether patients with laryngeal SCC treated with non-surgical organ preservation treatment would undergo SLT. Data was collected from the US National Cancer Database (NCDB); a total of 16,440 cases were analysed. Logistical regression analyses were performed on the same data for comparison. The ML model identified three main predictive factors for SLT: distance from residence to treating facility; days from diagnosis to start of treatment; and clinical T stage. The algorithm was able to predict SLT with an accuracy of 76%. While there was a high degree of concordance between the algorithm and regression analysis, there were some discrepancies. Overall, the authors maintain the preponderance of published evidence supports the model’s conclusions. The accuracy of the ML algorithm is limited by the quality of data and there are several limitations with the NCDB data, such as the low number of salvage laryngectomies compared to published literature. Additionally, the process by which an algorithm arrives at its conclusions may be impossible to know, resulting in a relative lack of transparency. Despite these limitations, ML has the potential to be a useful clinical tool in the provision of high quality, personalised cancer care.