This study looks at why college students keep using generative AI, like ChatGPT, in higher education. It examines factors such as how well the AI meets their expectations, their satisfaction with it, and the influence of their peers. The research was done in two stages, with online surveys completed by students from a local college in Hong Kong after one month and then again after five months of using ChatGPT. The data was analyzed using structural equation modeling (SEM). The results show that when the AI meets students’ expectations and they are satisfied with its performance, they are more likely to continue using it. However, the importance of performance expectations decreases over time. While social influence plays a role in initially getting students to use the AI, it doesn't significantly impact whether they continue using it. The study highlights the need to manage expectations, improve AI performance, and keep users satisfied to ensure long-term use of generative AI in education. These findings can help researchers and educators develop strategies to effectively and ethically incorporate AI tools in educational environments.