<p>Collective reflection plays a significant role in promoting pre-service science teachers’ professional development, yet encountering problems such as disorganization, superficial cognition, and intangible quality due to lack of teaching experience and critical ability. Therefore, it is necessary to scaffold the collective reflection among pre-service science teachers comprehensively. Given the potential of generative artificial intelligence (GenAI) in promoting teacher education, it was integrated into a scaffolded collective reflection for pre-service science teachers. This study employed epistemic network analysis (ENA) to trace the development process of pre-service science teachers’ collaborative discourse patterns in collective reflection. The results showed that the ChatGPT groups demonstrated more active on introducing multiple voices but neglected critical thinking and reasoning in collective reflection. Furthermore, the semi-structured interview results demonstrated that quite a few pre-service science teachers still had misconceptions on how to use GenAI in collective reflection. Suggestions are provided for improving pre-service science teachers’ collective reflection abilities in a GenAI-integrated environment.</p>