<p>Generative Artificial Intelligence (GenAI), as a promising emerging technology, has been widely applied across various fields. While existing research continues to explore the potential applications of GenAI in education, studies specifically focusing on the use of GenAI-augmented pedagogical agents (GPAs) in primary education remain limited. Based on the Hedonic-Motivation System Adoption Model (HMSAM), this study examined the application of GPAs and investigated students’ hedonic-motivation and behavioral intentions towards these agents. The Structural Equation Modeling (SEM) results from a survey of 330 primary school students confirmed the applicability of the HMSAM model. Except for the unverified positive effects of usability on perceived usefulness and curiosity on focused immersion, all other hypotheses were supported. The Artificial Neural Networks (ANN) results further revealed that joy is the most significant factor influencing students’ behavioral intentions to use the agents, followed by curiosity and perceived usefulness. Joy also serves as a key predictor of focused immersion, with control being the second most important factor. This study provides practical insights for educators and technology developers for optimizing GPAs to better serve students. Educational institutions are encouraged to provide diverse GenAI learning experiences to foster students’ curiosity and enjoyment; teachers can emphasize the real-world relevance of GPA use to strengthen students’ perceived usefulness; and GPA developers can incorporate learner feedback to better align agent functions with user needs. These findings offer guidance for the effective integration of GenAI technologies into educational practices.</p>