In this paper, we explore the development and effectiveness of a system that leverages LLM (GPT) to generate personalized learning advice based on individual learners’ progress and reflections accumulated online. By constructing prompts that incorporate both quantitative and qualitative data, GPT generated the advice. The appropriateness of the generated advice was evaluated from both “teacher” and “learner” perspectives. The results indicated that the advice closely matched what a teacher might offer. Additionally, implementing the advising system in actual classes and evaluating it through surveys showed that learners generally set their goals for the next week based on the system’s advice with a sense of satisfaction. Thus, it was found that the generated advice was generally appropriate from both “teacher” and “learner” perspectives.
