With the rapid development of generative AI, its deep integration into university students’ learning and daily life has made AI dependence increasingly prevalent. While existing studies have primarily focused on individual psychological mechanisms of AI dependence, they have paid less attention to the role of external social environments, particularly the increasingly prevalent phenomenon of academic involution in higher education. Guided by the I-PACE framework and integrating stress, coping theory, social comparison theory, and conservation of resources theory, we propose that academic involution atmosphere relates to students’ AI dependence through academic stress, with individual academic involution behavior moderating this association. Using PLS-SEM to analyze data from 500 Chinese university students, we find that academic involution atmosphere is associated with higher academic stress. Higher academic stress is associated with students viewing AI as a technological substitute, which correlates with greater AI dependence. Higher individual involution attenuates the positive association between involution atmosphere and academic stress. This moderated mediation model offers insights into the associations among these variables, providing insights for universities to alleviate stress, optimize the competitive atmosphere, and guide rational AI use.