The Double-Edged Effect of AI Use on Innovation Teaching Behavior among Primary and Secondary School Teachers in China: A Job Demands–Resources Perspective

Background As artificial intelligence (AI) technologies increasingly permeate the educational landscape, their impact on teaching innovation has attracted growing academic interest. While AI holds the promise of enhancing pedagogical practices, it also introduces stress-inducing demands, leading to a paradoxical "double-edged sword" effect. This study examines how AI use affects primary and secondary school teachers’ innovation teaching behaviors in China, employing the Job Demands–Resources (JD-R) model to investigate the dual pathways of resource gain and stress depletion, while also exploring the moderating role of school innovation support. Methods A cross-sectional survey was administered to 1,275 in-service Chinese teachers using validated scales for AI use, innovation teaching behavior, creative self-efficacy, work stress, and school innovation support. The hypothesized dual-path moderated mediation model was analyzed using the PROCESS macro in SPSS. Creative self-efficacy and work stress were tested as parallel mediators, while school innovation support was analyzed as a moderator of both mediation pathways. Results AI use significantly predicted innovation teaching behavior ( β = 0.047, p < 0.01). Creative self-efficacy served as a strong positive mediator (indirect effect=0.097, 69.86% of total effect), while work stress exerted a weaker, suppressive mediating effect (indirect effect = –0.005, −3.59%). School innovation support significantly moderated the stress pathway: Its presence exacerbates the negative impact of AI-induced stress on innovative behavior (moderated indirect effect = –0.008, 95% CI [–0.017, –0.002]). However, its moderating effect on the empowerment pathway via self-efficacy was non-significant. Conclusion This study provides empirical evidence for a dual-path mechanism through which AI use influences teaching innovation. While AI can empower teachers by enhancing creative self-efficacy, it simultaneously increases work stress that inhibits innovation. School innovation support does not significantly enhance teachers' creative self-efficacy, it may exacerbate the work stress induced by the AI use.These findings extend the JD-R model to AI-embedded educational settings, offering theoretical and practical insights for promoting sustainable teacher innovation during digital transformation.