The influence of ChatGPT on medical student learning: a systematic review of educational outcomes and cognitive impact

Aim. To evaluate whether the use of ChatGPT as a supplement to usual teaching improves medical students’ shortterm knowledge, clinical reasoning, and near-term performance. Materials and methods . We systematically searched PubMed, Scopus, ScienceDirect, SpringerLink, and Web of Science on 25 June 2025, for studies involving medical students that evaluated ChatGPT as a supplement to teaching and reported objective educational outcomes. Two independent reviewers screened records, extracted data, and assessed the risk of bias. A narrative synthesis was then conducted due to the level of heterogeneity in interventions and outcome measures across the studies. Results. Three randomized trials conducted in Germany, Turkey, and China met the inclusion criteria. ChatGPTsupported interventions improved or at least maintained short-term educational outcomes over the control groups for knowledge tests, clinical reasoning, and some of the Mini-Clinical Evaluation Exercise (Mini-CEX) domains. Structured and immediate ChatGPT feedback improved Clinical Reasoning Indicator-History Taking Inventory scores after a simulated patient encounter, and ChatGPT-generated explanations were not inferior to expert feedback in overall key-features question performance but were less effective for more complex items, where expert feedback remained superior. Overall, the risk of bias was judged to be low to some concerns, with likely unblinded Mini-CEX assessment noted as a significant limitation. Conclusion. ChatGPT used as a supervised adjunct to teaching showed value for short-term knowledge acquisition and clinical reasoning development.