<p>GenAI chatbots can be flexibly integrated into online learning platforms to create safer and more autonomous environments for inquiry-based learning. They enrich students’ learning experiences by offering personalized support and fostering deep inquiry through real-time feedback. However, students in chatbot-supported inquiry-based learning (CsIBL) environments require strong metacognitive abilities to regulate both inquiry processes and chatbot usage: they need to plan, monitor, and adjust their inquiry while critically evaluating chatbot responses and avoiding overreliance that may reduce active inquiry to passive reception. Without proper guidance, they often struggle to consistently apply metacognition for self-regulation. Metacognitive scaffolding has proven effective in fostering metacognition and self-regulated learning, yet few studies have examined its design and effects in chatbot-supported science education. To address this gap, this study designed a metacognitive scaffolding-integrated chatbot-supported inquiry-based learning (MS-CsIBL) approach to enhance students’ self-regulation and inquiry performance. A quasi-experimental design involving 85 eighth-grade students was employed. The experimental group (<i>N</i> = 43) adopted the MS-CsIBL approach, whereas the control group (<i>N</i> = 42) utilized the traditional CsIBL approach without scaffolding. Data were collected through surveys, learning logs, and interviews. The findings indicated that the approach significantly promoted the development of inquiry skills and supported positive learning experiences, as indicated by enhanced self-efficacy and a more favorable cognitive load profile. Moreover, it fostered a distinctive metacognitive behavior pattern centered on monitoring, where monitoring, evaluation, regulation, and reflection were closely interlinked. This study provides theoretical and practical implications for designing effective GenAI-supported scientific inquiry courses in K-12 education.</p>