As the field of artificial intelligence (AI) continues to progress, the use of AI-powered chatbots, such as ChatGPT, in higher education settings has gained significant attention. This paper addresses a well-defined problem pertaining to the critical need for a comprehensive examination of students' ChatGPT adoption in higher education. To examine such adoption, it is imperative to focus on measuring actual user behavior. While measuring students' ChatGPT usage behavior at a specific point in time can be valuable, a more holistic approach is necessary to understand the temporal dynamics of AI adoption. To address this need, a longitudinal survey was conducted, examining how students' ChatGPT usage behavior changes over time among students, and unveiling the drivers of such behavior change. The empirical examination of 222 Dutch higher education students revealed a significant decline in students' ChatGPT usage behavior over an 8 month period. This period was defined by two distinct data collection phases: the initial phase (T1) and a follow-up phase conducted 8 months later (T2). Furthermore, the results demonstrate that changes in trust, emotional creepiness, and Perceived Behavioral Control significantly predicted the observed change in usage behavior. The findings of this research carry significant academic and managerial implications, as they advance our comprehension of the temporal aspects of AI adoption in higher education. The findings also provide actionable guidance for AI developers and educational institutions seeking to optimize student engagement with AI technologies.