A New Era in Diabetes Management: Generative Artificial Intelligence

Diabetes mellitus (DM) is a rapidly increasing global health issue that requires effective selfmanagement to prevent complications and improve quality of life. In recent years, advancements in generative artificial intelligence (GenAI) have created new opportunities to support DM selfmanagement by providing personalized care solutions. This study is designed as a systematic review. Numerous studies in the literature have examined the contributions of GenAI models to DM self-management, and reviewing these studies is essential to provide a general framework on this topic. The primary aim of this study is to systematically examine research that utilizes GenAI in DM management. This systematic review was conducted in accordance with PRISMA guidelines. A comprehensive literature search was carried out between February and October 2024 across PubMed, Scopus, Web of Science, Google Scholar, Ulakbim, Türk Medline, and national databases. Using the keywords "diabetes," "generative artificial intelligence," and "diabetes self-management," studies published between 2018 and 2024 were identified. A total of 19 studies that met the inclusion criteria were analyzed in terms of the GenAI models used, application areas, and reported outcomes. Among the reviewed studies, GPT-based models were predominant, appearing in 53% of the research. In addition, models such as GAN, LSTM, WaveNet, GRU, Markov-Bayes, Google Bard, and Mobiguide were also utilized. Moreover, the findings of this study highlight that GenAI-based systems are widely adopted in DM selfmanagement and possess significant potential to facilitate this process. These systems not only provide information but also incorporate advanced support mechanisms that enhance patient monitoring and clinical decision-making processes. GenAI has made notable contributions to DM care, particularly by developing personalized care plans, offering tailored dietary and exercise recommendations, generating educational materials, predicting blood glucose (BG) levels, providing individualized guidance, and supporting clinical workflows. As GenAI continues to evolve and adapt to the specific contexts and demands of the medical field, its role in DM care is expected to become increasingly prominent. However, several challenges have been reported, including concerns over data security, privacy, misinformation generation, and suboptimal performance in detecting critical conditions such as hypoglycemia. Addressing these ethical, technical, and security-related limitations requires further research and technological advancements. Future studies should prioritize enhancing the reliability, usability, and diagnostic accuracy of GenAI applications to ensure their seamless integration into clinical practice.