Artificial intelligence (AI) and continuous glucose monitoring (CGM) are reshaping diabetes care by enabling glucose prediction, intelligent insulin delivery, personalized nutrition, and data-driven phenotyping. A new Chinese Medical Journal review highlights how AI unlocks CGM's full potential, moving beyond traditional blood sugar tracking toward proactive, individualized management — despite ongoing challenges in cost, privacy, and clinical integration.
Continuous glucose monitoring (CGM) devices, which measure interstitial glucose levels every 5–15 minutes through a subcutaneous sensor, generate vast longitudinal datasets far richer than traditional fingerstick tests. Yet these complex temporal measurements often exceed the analytical capacity of conventional statistical methods. Artificial intelligence (AI)—particularly machine and deep learning—offers a powerful solution to decipher these intricate patterns, detect subtle trends, and forecast glucose fluctuations before they become clinically significant.
In their review published in the Chinese Medical Journal on September 14, 2026, Professor Jian Zhou and colleagues from Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, China, along with researchers from Northeastern University, China, and the Chinese Academy of Sciences, China, systematically examine how AI-augmented CGM analytics are revolutionizing multiple aspects of diabetes care, including glucose prediction, closed-loop insulin delivery, clinical decision support, data-driven phenotypic clustering, and personalized nutrition.
AI-based models now forecast glucose levels with clinically acceptable accuracy for horizons of up to 1 hour or more, extending well beyond earlier proof-of-concept studies. Incorporating contextual factors, such as physical activity and psychological stress, further improves performance. These predictive capabilities are particularly meaningful given insulin's prolonged action in the body, enabling more anticipatory insulin delivery. When integrated into closed-loop “artificial pancreas” systems, such models promise to transition current hybrid systems—which still require manual meal announcements—into more intelligent and automated insulin management. Randomized clinical trials have already demonstrated that closed-loop systems improve time in range, reduce hypoglycemia, and lower glycated hemoglobin (HbA1c) across diverse age groups.
Beyond real-time control, AI-enabled clinical decision support systems analyze CGM data to guide individualized insulin adjustments for both type 1 and type 2 diabetes. Simulated and real-world evidence shows that these tools improve glycemic outcomes while also helping prioritize high-risk patients for review, thereby improving clinical efficiency. Moreover, unsupervised clustering of raw CGM data has uncovered distinct glycemic phenotypes that conventional measures like HbA1c often miss. This data-driven reclassification offers a novel framework for tailoring medication and lifestyle strategies to each patient's underlying pathophysiology.
Lifestyle interventions, especially nutrition, play a critical role in diabetes management. However, recent CGM analyses have revealed marked individual variability in postprandial glycemic responses to identical meals. Machine learning models that combine CGM profiles with diverse biological and behavioral data now accurately forecast individualized responses, enabling multimodal platforms and digital twin technologies to deliver tailored dietary and physical activity guidance. Early studies demonstrate that such approaches can reduce HbA1c and even support remission in type 2 diabetes.
Despite its promise, the integration of AI and CGM faces significant hurdles. High implementation costs and disparities in digital access risk exacerbating healthcare inequities. The sensitive nature of CGM data also demands robust ethical frameworks and legal safeguards to ensure privacy and algorithmic fairness across diverse populations. Equally critical are human and workflow considerations—AI tools must streamline, not disrupt, clinical practice to gain real-world acceptance. The authors emphasize that addressing these practical challenges is as important as enhancing predictive accuracy for successful implementation.
“The integration of AI and CGM is not merely a technological advancement but represents an inevitable evolution in diabetes care,” the authors conclude. “ It drives the transition from population-averaged management to individualized precision management, opening new frontiers for early risk detection and personalized treatment strategies."
***
Reference
DOI: 10.1097/CM9.0000000000004229
About Professor Jian Zhou from Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, China
Professor Jian Zhou is a Chief Physician, Professor, and Doctoral Supervisor at Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine. He currently serves as the Director of the Department of Endocrinology and Metabolism and Deputy Director of the Shanghai Diabetes Institute. He is a Shanghai Leading Talent and Shanghai Outstanding Academic Leader. His research focuses on the pathogenesis of diabetes, continuous glucose monitoring, and interventional studies. He has published extensively in journals including Diabetes Care, Diabetes , and Diabetologia . He serves as the Vice Chairman of the Diabetes Branch of the Shanghai Medical Association.
Funding information
This work was funded by the Noncommunicable Chronic Diseases–National Science and Technology Major Project (Nos. 2024ZD0532000 and 2024ZD0532001), the Shanghai Oriental Talent Program (Youth Project), and the Shanghai Leading Talent Program of Eastern Talent Plan (No. LJ2024121).
Chinese Medical Journal
Commentary/editorial
Not applicable
Artificial intelligence meets continuous glucose monitoring data: Genie in the bottle
14-Sep-2026