AI Attitudes and Reported GenAI Use Among Teacher-Education Students at a Romanian University: Associations with Perceived Extraneous Load and Perceived Learning Enhancement

Generative artificial intelligence (GenAI) can support varied academic practices, but favourable attitudes and perceived benefits do not establish how students use it or what they learn. This cross-sectional study examined associations among AI attitudes, reported GenAI use, and perceived cognitive experiences in 472 students enrolled in teacher-education programmes at one Romanian university. The analytical specification was documented after the data had been examined but before the final regression analyses; the study was not preregistered. Four ordinary least squares models used heteroscedasticity-consistent (HC3) inference and family-specific Benjamini–Hochberg adjustment. Positive AI Attitudes were associated with both Strategic Use (β = 0.518) and Shortcut Use (β = 0.251; both q < 0.001), whereas Reverse-Scored Negative AI Attitudes did not meet the focal FDR criterion for either use index. Neither use index was clearly associated with Perceived AI-Related Extraneous Load. In exploratory M4, Strategic Use was positively associated with perceived improvement in understanding (β = 0.199, q < 0.001), whereas Shortcut Use was inconclusive (β = 0.026, q = 0.522). The findings suggest that general AI attitudes, reported practices, and perceived learning benefits should be treated as distinct aspects of students’ experiences with GenAI. Given the limited psychometric support for some observed indices, the absence of preregistration, and the cross-sectional design, the results do not establish causal effects, changes in student engagement, or independently assessed learning.