Generative AI and multimodal pedagogy: implications for teaching, learning, and assessment

Generative AI tools have made multimodal content creation accessible to teachers, sharply reducing the time and expertise needed to produce it—a shift with significant pedagogical implications. This article examines what this democratisation means through the lens of multimodal pedagogy, which treats teaching as the deliberate orchestration of multiple semiotic modes and draws on several theoretical traditions, most prominently social semiotics and the cognitive theory of multimedia learning. Two interconnected arguments organise the analysis. First, generative AI lowers the production barrier for multimodal educational materials to a degree that reconfigures what is practically feasible in classrooms and course design, extending possibilities that interactive digital media began opening two decades ago. Second, this expanded capacity amplifies rather than substitutes for pedagogical expertise: without sound design grounded in how learning works, more powerful tools risk producing more elaborate noise. Drawing on the multiliteracies tradition, Mayer's multimedia learning principles, and Universal Design for Learning, the article maps four functional categories of AI tools against three evaluative criteria—pedagogical completeness, cognitive soundness, and adaptive flexibility—and discusses their implications for assessment, teacher preparation, and ethical governance. The analysis acknowledges the still-crystallising nature of this domain: empirical evidence for multimodal pedagogy's superiority over text-based instruction remains thin, AI-generated content carries structural limitations in accuracy and bias, and most claims about learning gains await rigorous validation. The article aims to provide a theoretically grounded orientation for educators, course designers, and researchers navigating a rapidly evolving intersection of AI capability and pedagogical principle.