Large Language Models and Vision-Language Models in Gastroenterology and Hepatology: A Scoping Review of Clinical Tasks, Enhancement Strategies, and Evaluation Frameworks

This registration describes the protocol for a scoping review mapping empirical research on large language models (LLMs) and vision-language models (VLMs) in gastroenterology and hepatology. The review will search PubMed, Embase, Web of Science and Scopus from January 2023 onward, supplemented by citation chasing, and will include peer-reviewed original empirical studies published in English or Chinese. It asks how the evidence is distributed across digestive subspecialties, clinical tasks and study designs; which enhancement strategies and reference standards are used; how far evaluation reaches along the path from technical performance to patient outcomes; and which safety constructs remain unmeasured. All included studies will be charted at a common Tier-1 level, with a nested Tier-2 deep extraction for studies meeting pre-specified clinical or safety criteria. The protocol is registered before any database search is executed.