GenAI in nutritional sciences (GAINS): A systematic review and reporting framework for future research

Area of Science:

• Nutritional Sciences
• Artificial Intelligence in Healthcare
• Medical Informatics

Background:

• Large language models (LLMs) are increasingly integrated into healthcare.
• LLM-powered chatbots are being explored for their potential to offer nutritional guidance.
• Existing evidence on the quality of chatbot-provided nutritional advice requires synthesis.

Purpose of the Study:

• To systematically review and synthesize evidence on the quality of nutritional advice from LLM-powered chatbots.
• To assess the suitability of chatbots for advice on metabolic diseases, food allergies, nutrient intake, and pregnancy/lactation nutrition.
• To identify limitations and methodological gaps in current research.

Main Methods:

• A pre-registered systematic review (CRD42025619448) was conducted.
• An extensive search identified 2469 records, with 13 studies meeting inclusion criteria.
• Standardized data extraction and quality evaluations were performed.

Main Results:

• LLM-powered chatbots show potential but struggle with complex cases and may provide incorrect information.
• Substantial methodological heterogeneity was found in evaluation criteria and study design.
• Limitations include subjective measures and lack of reproducibility testing.

Conclusions:

• LLM-powered chatbots demonstrate potential for supporting nutritional advice but are not yet suitable for unsupervised use.
• Significant methodological gaps necessitate improved rigor, transparency, and comparability in future studies.
• A proposed reporting guideline, Generative AI in Nutritional Sciences (GAINS), aims to address these gaps.