A Narrative Review of the Applications of Artificial Intelligence in Diabetes Patient Education

Abstract

Context: Given the increasing prevalence of diabetes and its substantial health consequences, artificial intelligence (AI) has attracted growing attention as an innovative approach to patient education and self-care management. However, comprehensive evidence regarding the role of AI in diabetes-related education and care remains limited. This narrative review examined the applications, outcomes, and challenges of AI in patient education for diabetes. Evidence Acquisition: This narrative review identified relevant articles through searches of PubMed, Scopus, Web of Science, Springer, and Google Scholar. Studies published between 2015 and 2025 were considered, with particular emphasis on publications from the past five years. Search terms included "artificial intelligence," "self-management," "chatbots," "mobile applications," "patient education," and "diabetes." Studies were included if they directly addressed AI technologies used for diabetes patient education or self-management. Results: AI-based interventions were categorized into four main groups: 1) chatbots and virtual assistants, 2) mobile applications, 3) intelligent educational systems, and 4) large language models. These technologies demonstrated the potential to improve diabetes education, support self-management behaviors, and enhance patient engagement. Conclusions: AI-based educational tools show promising potential to support diabetes education, self-management behaviors, patient engagement, glycemic outcomes, and treatment adherence through personalized education, continuous interaction, and feedback. However, findings remain heterogeneous, the overall quality of the evidence has not been systematically evaluated, and conclusions should therefore be interpreted with caution. Data privacy concerns, the lack of standardized evaluation frameworks, limited accessibility, and the need for professional supervision remain important barriers to widespread implementation.