Patterns of Generative AI Use in Science Teaching: A National Survey of K–12 Teachers

While generative artificial intelligence (GenAI) has been widely discussed in science education research, there lacks grounded understanding of how science teachers are using these tools in practice. This study examines the use of GenAI in K–12 science teaching through a national survey of U.S. teachers. Drawing on a probability-based sample of 163 science teachers in April 2025, we analyzed their patterns of adoption, instructional use, and perceived impact. About 66% of science teachers reported using GenAI at all, with most being emerging users who use these tools weekly or monthly. Teachers primarily used GenAI for instructional planning, while substantially fewer used it to support student-centered learning with GenAI. Latent class analysis identified three user profiles including planning-oriented users, in-class users, and extensive users. The majority of teachers concentrated on planning-focused uses, while a small group reported teaching with and about GenAI as their major use. Teachers reported limited instructional impact of GenAI, which was seen more as a tool to improve productivity than to transform pedagogy. Latent profile analysis revealed variation in the perceived impact. No significant differences in use or perceived impact were observed across teacher or school characteristics. Findings revealed a gap between the potential of GenAI to transform science education and its current role as an efficiency-oriented tool, underscoring the need for professional learning and instructional design that support integrating GenAI into core instructional practices and student learning.