Quality assessment of ChatGPT-generated medical information for healthcare professionals and patients: a systematic reviewv

Abstract

ChatGPT, an artificial intelligence (AI)-based language model, has gained popularity in medicine, although concerns remain regarding the accuracy and clinical applicability of its generated content. In accordance with PRISMA guidelines, a systematic review was conducted using the PubMed, EBSCO, and Scopus databases, including 29 studies published between 2022 and 2024. The review included publications evaluating the accuracy, readability, comprehensiveness, and clinical utility of medical information generated by ChatGPT. ChatGPT’s content was generally consistent with clinical knowledge, although the quality varied depending on language, query formulation, and medical domain. Responses in English were usually more accurate. Limitations included lack of references, insufficient contextual and emotional understanding, and inconsistencies with medical guidelines in complex cases. ChatGPT shows potential as an educational and clinical decision-support tool but requires further refinement and standardisation of evaluation methods.