Predicting Flow State and Learning Engagement in <scp>GAI</scp> ‐Powered Virtual Influencer‐Mediated Informal Foreign Language Learning

ABSTRACT This study examines the role of Generative Artificial Intelligence (GAI)‐powered virtual influencers in enhancing foreign language learning engagement within informal digital environments in China. Drawing on the Hedonic‐Motivation System Adoption Model, we develop and test a theoretical framework that incorporates perceived ease of use, perceived usefulness, joy, perceived behavioural control, parasocial relationships and flow state. Using structural equation modelling to analyse survey data from 368 Chinese college English learners engaged with GAI‐powered virtual influencers on Bilibili, our findings demonstrate that perceived behavioural control emerges as the most significant predictor of both flow state ( β = 0.342, p &lt; 0.01) and learning engagement ( β = 0.524, p &lt; 0.001), as well as the importance of perceived ease of use. The study reveals that parasocial relationships ( β = 0.177, p &lt; 0.01) and joy ( β = 0.155, p &lt; 0.01) also significantly contribute to engagement levels. These results advance our understanding of technology‐mediated foreign language learning by highlighting the unique mechanisms through which GAI‐driven virtual influencers facilitate learning in informal settings. The research provides practical implications for designing educational technologies that effectively combine entertainment and pedagogical value to optimise learner engagement.