Integrating artificial intelligence (AI) at universities opens up new avenues for accessing and acquiring contents and methodological knowledge, as well as for managing academic work. We investigate the use of ChatGPT, an AI-powered chatbot, in the context of physics studies. Specifically, we focus on experiences in the exercise course for the introductory lecture on Theoretical Mechanics in the Physics Bachelor study program. Aspects beyond this course are addressed as well. Our approach includes additional exercises supporting the lectures that illustrate to students the potential and current limitations of ChatGPT in its freely available version. For this purpose, we used two different types of tasks. On the one hand, this involved using ChatGPT to solve computational exercises, and on the other hand, utilizing the chatbot for conceptual explanations of physical phenomena, definitions, and everyday examples. Additionally, we conducted an online survey among the students. The results show that students found the provision of conceptual explanations and examples by ChatGPT to be helpful. However, the support for computational tasks in the freely available version of ChatGPT at that time was perceived as unreliable and weak. Our framework contributes to the discussion on the meaningful use of AI technologies in university teaching and provides insights into students’ perceptions and use of these technologies.