The potential of classroom videos to enhance reflective practices in pre-service teacher education is hindered by the sheer volume of captured activities. An AI-powered teacher dashboard could address this challenge by analyzing and visualizing information extracted from these videos, supporting reflection in video-based professional learning contexts. To explore how this dashboard affects the depth of cognitive and critical reflection among pre-service teachers, we conducted a quasi-experimental study with 48 pre-service teachers from a university in eastern China. All pre-service teachers in each group watched the same classroom videos and discussed lesson plan improvements using the QQ chat tool. In the implementation condition, 24 pre-service teachers used the AI-powered teacher dashboard, while the contrast condition (n = 24) did not. Analysis of collaborative discourse, alongside descriptive and epistemic network analysis, revealed that the AI-powered teacher dashboard improved the quality of their reflection. These findings offer theoretical and practical insights for future research and practice in AI-powered video-based reflection approaches in teacher education and professional development.