Teachers’ nonverbal communication, such as facial expressions, posture, and paralanguage, significantly influences students’ motivation and academic outcomes. These elements, collectively referred to as nonverbal teaching skills in this study, are essential for pre-service teachers to master. Although personalized feedback on such skills enhances skill development during microteaching, providing timely personalized feedback is labor-intensive for teacher educator. Artificial intelligence-based automated feedback offers a scalable solution, yet its effectiveness compared to traditional human feedback remains underexplored. This study addresses this gap through a randomized controlled experiment involving 67 pre-service teachers in Mainland China, assigned to three conditions: human feedback (n = 20), automated feedback (n = 21), and a hybrid approach combining both (n = 26). Five indicators of nonverbal teaching skills were analyzed across four teaching sessions. Key findings revealed that: (1) all groups exhibited an increased frequency of posture changes and spent more time facing forward across sessions; (2) the human feedback group experienced a reduction in facial expression variability, while the other groups showed no changes; and (3) no significant changes occurred in speech disfluency or positive facial expressions in any group. These results underscore artificial intelligence-based feedback as a viable complement to human feedback in teacher education.