Generative artificial intelligence has quickly become part of students' academic practice, yet adoption depends on more than awareness or access to a chatbot. This study examined the factors associated with behavioural intention and actual academic use primarily among bachelor-level students among different colleges in Kathmandu, Nepal. A cross-sectional online survey produced 74 consenting submissions. Six uniform response patterns across all 20 scale items were excluded through a pre-specified data-quality rule, leaving 68 records for analysis. Perceived usefulness, perceived ease of use, facilitating conditions, behavioural intention, and actual use were measured with four-item scales; institutional guidance was measured with one teacher-clarity item. Cronbach's alpha ranged from.792 to.939. The HC3-robust regression predicting behavioural intention was significant, F(4, 63) = 43.79, p <.001, and explained 73.5% of the variance. Facilitating conditions were the strongest independent predictor (Beta =.379, p <.001), followed by perceived ease of use (Beta =.334, p =.004) and perceived usefulness (Beta =.278, p =.020). Institutional guidance was not significant after the other predictors were controlled (Beta = -.026, p =.776). Behavioural intention strongly predicted actual use (Beta =.629, p <.001), explaining 39.5% of its variance. Four of the five hypotheses were supported. The findings indicate that adoption is most likely when students perceive academic value, can use the tools without excessive effort, and have adequate devices, access, help, and institutional support. The single guidance item should not be interpreted as evidence that policy is unimportant; a broader multi-item measure is required. The results are specific and are not representative of all students in Kathmandu Valley or Nepal.