The arrival of ChatGPT has sparked a broad debate about the future of education in the context of generative artificial intelligence (GenAI). Initial reactions ranged from enthusiasm to fear, but recent research suggests that, assuming ChatGPT has the potential to be a useful teaching tool, it could play a constructive role in the classroom if used with a clear educational purpose. In this article, I analyze the educational role that ChatGPT should play in different learning modalities, considering the latest empirical studies in higher education. Based on a critical synthesis of the current literature, I consider how students and teachers interact with ChatGPT and review different types of usage profiles, ranging from behaviors derived from automated practices to those of collaborative teaching and reflection. I examine the tensions between the tool's efficiency and deep learning, and I argue that its integration requires technical training, but above all, broader AI literacy that considers ethical, critical, and metacognitive aspects. It also examines the extent to which the arrival of ChatGPT challenges course design and traditional forms of assessment, forcing us to redefine the concepts of originality, authorship, and assessment in education. The argument demonstrates that ChatGPT's instructional mediation is more valuable than its outcome. I argue that institutions should not prevent its use; instead, they should establish rules and regulations that encourage responsible, fair, and useful use. In conclusion, it seems to me that ChatGPT should not be considered a substitute for teaching, but rather a way to rethink our education, give students more freedom, and enhance the learning experience.