The complexity and volume of data in healthcare entail that artificial intelligence (AI) and associated technologies are becoming an essential component of life sciences. The scientific literature explores the benefits, contexts of application, ethical implications, and future developments. In order to 1) identify the topics characterizing the literature on AI in the clinical domain, 2) detect semantic categories, and 3) validate them, a methodological approach based on the combination of natural language processing and machine learning for classification was performed. Two main semantic categories were identified: diagnostics and treatment, which are used to manually annotate each document. Finally, we tested our semantic classification through machine learning. The findings suggest clear differences between the two categories, mainly based on AI-assisted meta-analyses and clinical decision support systems, with just a quota of scientific papers encompassing both semantic pillars. This proportion of documents is pivotal to changing the semantic classification.