Despite advances in AI pedagogical agents, research on developing creative instructional design remains limited, and how they affect pre-service teachers’ creative thinking—especially with metacognitive scaffolding—is unclear. Based on metacognitive and creativity investment theories, this study examines how such scaffolding in an AI agent fosters creative instructional design among pre-service preschool teachers, and whether critical thinking mediates this relationship moderated by AI dependency. A quasi-experimental design was used with 120 pre-service preschool teachers, who were assigned to the experimental group or the control group. Data on metacognitive awareness, AI dependency, critical thinking, and creative thinking were gathered through valid measuring instruments, and innovative curricula were evaluated by experts. The results show that the experimental group achieved much better scores on creative instructional design ideas after the test than the control group. The moderated mediation analysis revealed a critical thinking-mediated pathway that was moderated by AI dependency. In conclusion, AI pedagogical agents with metacognitive scaffolding (MAS-based) improved critical thinking and promoted deeper, more independent creative thinking, thus improving creative instructional design, through a pathway that is moderated by the degree of AI dependency. This study offers valuable theoretical and practical insights to cultivate creative teaching skills.
Zhang et al. (2026) studied this question.