Chat, Gemini and Claude at the dinner table: assessing general-purpose AI tools for carbohydrate counting in the context of type 1 diabetes
Authors' conclusion
General-purpose AI tools can provide carbohydrate estimates but show variable accuracy. They may complement, not replace, structured education in T1D management.
Study context
- Population
- Thirty meals of varying type and complexity, in the context of T1D carbohydrate counting
- Participants
- 30 meals
- Comparison
- ChatGPT, Gemini and Claude vs expert dietitians for carbohydrate estimation
- Outcome
- Absolute error vs reference carbohydrate content
Methodological detailsmixed-effects regression with meal as random effect, standardized prompt
- Mixed-effects regression with meal as random effect
- Standardized prompt
Performance
As reported in the abstract
Secondary outcomes
Large overestimations (≥+20 g)
Performance by meal type or complexity
- Performance did not vary by meal type or complexity
Original publication
Chat, Gemini and Claude at the dinner table: assessing general-purpose AI tools for carbohydrate counting in the context of type 1 diabetes. Diabetes Res Clin Pract. Published 26 November 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →