At present, Artifi cial Intelligence (AI) is deeply reshaping the landscape of high school geography education. While technology off ers clear advantages in making teaching processes more effi cient and expanding its reach, the specifi c risks within the subject have not been fully examined. Therefore, this study develops a risk assessment system that covers thinking development, data collection, ethics, and social impact, using the Entropy Weight Method for quantitative analysis. Our regression analysis shows that how people use technology is the main factor causing these risks. Furthermore, a case study on "Three Lanes and Seven Alleys" confi rms that AI has limits when dealing with complex human-environment relationships and teaching geographic values. By analyzing these risks through data, this paper aims to provide a guide for the smart transformation of geography teaching. We hope to ensure that while we embrace innovation, we also stay true to the core principles and educational purposes of the subject.