Large language models in obesity: a systematic review

Generative artificial intelligence (AI), particularly large language models (LLMs), is rapidly evolving and holds significant potential for addressing the multifaceted challenges of obesity management. This systematic review synthesizes current research on LLM applications within the obesity field, critically evaluating their performance, limitations, and key future research directions.

Electronic databases, including MEDLINE, OVID, and SCOPUS, were systematically searched from inception to August 2025 to identify eligible studies. Eligible studies investigated the potential role of LLMs in obesity medical and surgical care. Key findings were extracted and synthesized narratively. The ROBINS-I tool was used to assess the risk of bias across each study.

Thirty-three studies met the inclusion criteria. The analysis reveals that LLMs are being applied across a diverse range of obesity-related topics, including personalized nutrition, educational interventions, guideline-directed medical therapy, weight loss strategies, anti-obesity medication information, and motivational interviewing. While some LLMs demonstrated promising accuracy and utility, substantial variability was observed. Numerous studies highlighted limitations, including inconsistent recommendations, inaccuracies, difficulties in handling complex scenarios, and a potential for biased outputs.

Generative AI and LLMs show considerable promise for enhancing various aspects of obesity management, from personalized interventions to clinical decision support. However, the current state of the technology exhibits crucial limitations regarding accuracy, consistency, and the ability to handle nuanced clinical situations, mandating a continued critical role for clinician oversight and validation. Further research is imperative to address these shortcomings, focusing on improving model training, validating performance in real-world settings, and addressing ethical considerations before widespread clinical implementation.