This study explores how Generative AI (GenAI) tools—ChatGPT, Bing Copilot, and Google Gemini—impact learning and engagement in university-level physics education. Merging systematic literature review and classroom experimentation, we evaluate how GenAI supports conceptual understanding, problem-solving, and user experience. Our findings indicate that GenAI enables flexible, personalized learning by simplifying complex topics and providing real-time feedback. Distinct affordances emerged across platforms: ChatGPT for logical reasoning, Bing Copilot for procedural tasks, and Google Gemini for visual explanations. However, limitations persist, including AI’s inability to address deep conceptual misconceptions and the risk of over-reliance. From an HCI perspective, this work underscores the importance of designing AI-powered educational tools that encourage active learning while aligning with human cognitive needs. We argue that careful integration and instructor mediation are critical for maximizing learning outcomes. This research contributes to the design of human-centered AI systems that enhance STEM education through adaptive, responsive, and interactive learning environments.