<p>Generative artificial intelligence (GenAI), such as ChatGPT, has been increasingly employed to enhance educational practices. Despite its promising applications, few studies have explored its specific use in STEM education. This study investigates the impact of ChatGPT-assisted STEM lesson plan design activities on the quality of lesson plans created by preservice teachers and explores their experiences with the ChatGPT-assisted process. We utilized explanatory mixed-methods design, comprising a quantitative quasi-experimental design and qualitative semi-structured interviews. In the quantitative phase, both the experimental and comparison groups consisted of 23 preservice primary teachers. The experimental group engaged in ChatGPT-assisted STEM lesson plan design activities, while the comparison group participated in similar activities without ChatGPT support. The qualitative phase involved nine participants from the experimental group. We used a STEM lesson plan evaluation form and a semi-structured interview form for data collection. The quantitative results indicated that activities significantly improved the quality of STEM lesson plans created by preservice teachers in both groups. However, the experimental group demonstrated more substantial improvements than the comparison group. The qualitative results revealed that these improvements might be due to the effective collaboration between preservice teachers and ChatGPT in several areas, such as generating ideas, creating questions, and suggesting teaching techniques. Besides, ChatGPT fostered preservice teachers by promoting creativity, connecting lessons to real-life situations, and integrating multiple STEM disciplines into their lesson plans. This study provides evidence that integrating GenAI, specifically ChatGPT, into preservice teacher education has great advantages for preparing future teachers for their professional roles.</p>