Empowerment or burden: promoting university students’ sense of gain in generative AI application based on the CAB-O framework

Generative AI (GenAI) has raised questions regarding how students’ reliance on AI support influences their learning experiences and perceived gains. Drawing on the CAB‑O framework, this study examined factors associated with university students’ sense of gain in AI-assisted learning. A survey of 850 students with GenAI experience from 14 Chinese universities was analyzed using structural equation modeling (SEM). Results showed that AI literacy positively predicted both content‑based and experience‑based gain, while unexpectedly increasing AI usage anxiety, which was positively associated with experience‑based gain. Willingness to communicate and AI dependency demonstrated significant positive associations with both gain dimensions, with AI dependency showing the strongest effects. AI usage shame showed no significant association with content‑based gain and only a marginal negative association with experience‑based gain. Multi‑group analysis revealed that STEM students reported higher levels of AI literacy, willingness to communicate, and content‑based gain than Non‑STEM students, whereas no significant differences were found in affective variables or structural pathways. These findings challenge the view that AI dependency is inherently maladaptive and highlight the importance of fostering critical AI literacy and reflective GenAI engagement.