Evaluating ChatGPT and DeepSeek as AI-Powered Supports for Physics Teachers

The rapid emergence of generative artificial intelligence (AI) is transforming educational practice, offering new opportunities for dialogic interaction, structured explanation, and adaptive feedback. This paper examines the pedagogical potential of two natural language processing systems, ChatGPT and DeepSeek, in the context of physics education, with a particular focus on conceptual understanding and instructional design. Drawing on constructivist, behaviorist, and sociocognitive perspectives, as well as the emerging framework of AI literacy, we analyze how these tools align with different phases of the instructional cycle. ChatGPT supports inquiry and conceptual change through adaptive, dialogic interactions that surface and address students’ intuitive models. DeepSeek emphasizes semantic precision and structured explanations, supporting formal reasoning and curricular alignment. Empirical studies indicate that these affordances are complementary: when integrated strategically, they can foster both exploratory engagement and disciplinary rigor. However, their effective use depends on teacher mediation, critical engagement, and thoughtful sequencing within instructional frameworks such as the 5E model. The paper concludes by identifying key limitations in the current research, including the scarcity of longitudinal studies, issues related to epistemic agency, and the need for teacher professional development. Together, these insights contribute to a nuanced understanding of how AI can be integrated meaningfully into physics education.