Key points are not available for this paper at this time.
Background With generative AI increasingly embedded in higher education, understanding the factors that shape students' adoption of AI-driven learning assistants has become essential. Drawing on the UTAUT2 framework, this study investigates Chinese university students' acceptance and actual use of DeepSeek, an open-source general-purpose large language model. Methods Survey data from 480 students were analyzed using confirmatory factor analysis and structural equation modeling. Results The results indicate that performance expectancy, facilitating conditions, price value, and habit significantly predict behavioral intention, while behavioral intention, facilitating conditions, and habit predict actual usage. Habit emerged as the strongest determinant of both intention and use, highlighting the central role of routinization in sustaining engagement with educational AI. Conclusion These findings highlight the importance of routinization in sustaining engagement with educational AI. The study extends the UTAUT2 framework to the context of large language models and offers practical implications for universities and developers in promoting both initial adoption and continued use.
Xu et al. (Fri,) studied this question.