What is it about?
People with type 1 diabetes often struggle to estimate carbohydrates in their meals — a key step for insulin dosing. This study tested two advanced AI models (ChatGPT-4 and Gemini) in everyday settings: social dinners, restaurant meals, and home cooking. AI accuracy improved when more meal details were available, suggesting a potential role in diabetes care with further validation.
Featured Image
Photo by Solen Feyissa on Unsplash
Why is it important?
This is the first real-world comparison of generative AI tools for carbohydrate counting in type 1 diabetes. The study highlights the potential of large language models to support patients in everyday life and contribute to digital health innovation.
Perspectives
As a physician and researcher in endocrinology, I’ve seen firsthand how difficult it can be for people with type 1 diabetes to estimate carbohydrates — especially in real-life settings like restaurants or social events. I wanted to test whether generative AI could offer real, practical help. I was surprised by how well these tools performed when given enough detail. I believe this could be the start of a new era in digital diabetes support.
Nicola Tecce
Universita degli Studi di Napoli Federico II
Read the Original
This page is a summary of: AI-Powered Carbohydrate Counting for Type 1 Diabetes: Accuracy and Real-World Performance, Diabetes Care, May 2025, American Diabetes Association,
DOI: 10.2337/dc25-0303.
You can read the full text:
Resources
Contributors
The following have contributed to this page
