Artificial intelligence (AI) in nutrition: A case-based comparison of generative AI models.

Area of Science:

• Medical Informatics
• Artificial Intelligence in Medicine
• Clinical Nutrition

Background:

• Clinical nutrition (CN) is increasingly complex due to chronic illness and malnutrition.
• Physician education in CN is limited in the US.
• Large language model (LLM)-based artificial intelligence assistants (AIAs) are emerging tools for clinical decision-making.

Purpose of the Study:

• To evaluate the performance of four LLM-based AIAs in complex clinical nutrition cases.
• To assess the clarity, relevance, evidence, and clinical utility of AIA responses by CN experts.

Main Methods:

• Retrospective evaluation of four LLM-based AIAs (ChatGPT, OpenEvidence, Gemini, Copilot).
• Utilized five complex clinical nutrition cases.
• Responses were blinded and reviewed by five CN physician experts using an eight-item assessment tool.

Main Results:

• All AIAs provided clinically appropriate responses.
• Gemini achieved the highest scores for relevance (4.04) and clarity (4.16).
• Overall satisfaction ranged from 3.08 (Copilot) to 3.84 (Gemini); citation quality and originality were limited.

Conclusions:

• LLM-based AIAs can reliably replicate expert reasoning in clinical nutrition.
• AIAs show potential in supporting physicians lacking specialized CN expertise, addressing knowledge gaps.
• Future applications may include AIA-enabled e-consultation for nutrition education.