AI-enhanced wearables are rapidly reshaping how people manage Type 2 diabetes and prediabetes. A new meta-review from the University at Buffaloprovides the clearest picture yet of what these devices can offer, and what must still be resolved before they become routine in diabetes care. It is the first comprehensive analysis to consolidate findings from thousands of studies on AI-driven wearables, including continuous glucose monitors (CGMs), activity trackers and multimodal biosensors.
Lead researcher Dr. Raphael Fraser explains that AI fundamentally changes the role of glucose monitoring. Traditional CGMs provide a few readings per day, while AI-enhanced CGMs deliver continuous insights, flagging trends and even predicting glucose shifts 60 to 120 minutes ahead. “AI turns CGMs from a rear-view mirror into a heads-up display,” Fraser notes. This shift supports more proactive care, helps patients avoid extreme glucose fluctuations, and may reduce progression from prediabetes to diabetes. The research was published in npj Digital Medicine.
AI-enabled wearables show potential
Across the 60 high-quality studies reviewed, AI-enabled wearables consistently showed potential to improve glycaemic control, personalize guidance based on behaviour and lifestyle, and relieve clinicians by filtering large data streams. Yet the review also highlights major gaps.
Many AI systems still function as “black boxes,” offering alerts without explaining the underlying cause, making it difficult for patients to act with confidence and clinicians to trust recommendations. Small sample sizes, inconsistent datasets and limited demographic diversity in research further restrict generalizability.
Technical challenges also remain. Different AI models, ranging from LSTM networks to transformers, vary widely in performance, interpretability and suitability for continuous biosensor data. The most clinically promising systems are those that integrate multiple data types (glucose, sleep, heart rate, physical activity) while remaining explainable and transparent.
Several issues remain
The researchers emphasize that before AI-enhanced wearables can become standard tools in diabetes management, developers must address issues around model transparency, data quality, ease of integration into clinical workflows and the cost and availability of devices.
Still, the direction is clear: AI-driven wearables have the potential to transform diabetes care by enabling earlier detection of risk patterns, personalized coaching and more efficient clinical decision-making. As Fraser puts it, “These tools could fundamentally shift diabetes care from reacting to problems to preventing them.”
Wearables and diabetes
Earlier this year, an international study showed that smartwatches combined with a supportive health app can significantly improve exercise adherence in people recently diagnosed with type 2 diabetes. Participants followed a home-based exercise program and received guidance through a virtual coach. By linking smartwatch activity data to the app, the program provided biofeedback and personalized recommendations to help users meet the goal of 150 minutes of physical activity per week.
The MOTIVATE-T2D study reported an impressive 82% retention rate after 12 months. Participants using the smartwatch-app combination exercised more consistently and saw measurable health benefits, including improved blood sugar control, lower systolic blood pressure and better cholesterol levels.
