Cohort Differences in University Students’ AI Readiness, Anxiety, and Adoption Intention: A Comparison of the 2022 and 2024 Survey Cohorts

The diffusion of generative artificial intelligence (AI) has raised questions about university students’ readiness to use it. We conducted a secondary analysis of 1205 records from 1146 students across eight countries and three survey waves (2022–2024). Fifty-nine students participated in both the 2022 and 2024 waves; analyses therefore used student-clustered inference. The primary comparison involved Slovak information-technology undergraduates surveyed in 2022 (186 records, all collected before the 30 November 2022 public release of ChatGPT-5) and 2024 (180 records). The 2024 cohort reported higher behavioural intention (d = 0.31; Holm-adjusted p = 0.024) and lower self-rated AI literacy (d = −0.36; Holm-adjusted p = 0.004); anxiety was higher only in supporting multi-country and latent analyses. Equality constraints produced little additional deterioration in the invariance models, but the imperfect configural fit, including full-model CFI = 0.812 and TLI = 0.796 and weak behavioural intention fit, limits the strength of this evidence. AI literacy additionally had α = 0.69, AVE = 0.48, and only partial scalar invariance, so its cohort difference is exploratory. Discriminant validity was mixed: HTMT exceeded 0.85 for relevance–career motivation (0.872) and intrinsic motivation–satisfaction (0.856), but no pair exceeded 0.90 and prespecified collapsed-factor models fitted worse than the ten-factor model. The principal standardised multiple regression explained 65.2% of behavioural-intention variance; demographic, educational, national, and temporal adjustment (R2 = 0.681) and parsimonious nonlinear sensitivity analyses preserved the central coefficient pattern. These repeated cross-sectional results describe cohort differences and associations, not causal effects of ChatGPT or any other tool.