["Journal of Computer Assisted Learning, Volume 42, Issue 5, October 2026. ", "\nABSTRACT\n\nBackground\nRapid adoption of artificial intelligence (AI) in education is reshaping teaching, learning, and administration. Beyond technical skills, AI integration requires cognitive, metacognitive/self‐regulatory, and epistemic preparedness; thus, latent barriers may constrain how secondary school principals interpret, evaluate, and use AI‐generated information in decision‐making, creating a need for a psychometrically sound instrument.\n\n\nObjectives\nThis study aimed (1) to develop and (2) to validate the AI Learning, Metacognitive, and Epistemic Barriers Scale (AIME‐B) for secondary school principals in Jordan, to assess cognitive, metacognitive/self‐regulatory, and epistemic constraints that hinder reflective, context‐appropriate engagement with AI outputs in school administration and decision‐making.\n\n\nMethods\nUsing an exploratory mixed‐methods design, data were collected from 712 secondary school principals in Jordan (2025). The initial psychometric form contained 40 items; following item purification, the final scale retained 38 items loading on six factors.\n\n\nResults and Conclusions\nInternal consistency was high (α = 0.891–0.958; ω = 0.880–0.958), and within‐model convergent evidence was acceptable (AVE > 0.50; CR > 0.84). The first‐order CFA supported six correlated dimensions. Revision‐stage comparisons using the item‐level confirmatory dataset favoured the correlated six‐factor model over one‐factor, four‐factor, two theoretically plausible five‐factor, second‐order, and bifactor alternatives. The bifactor solution did not support a strong general factor, reinforcing primary interpretation of the six subscales. Gender invariance and bootstrapped EGA provided supplementary evidence regarding score comparability and clustering stability. Overall, the AIME‐B provides initial, context‐specific evidence supporting the internal structure and internal consistency of scores among Jordanian secondary school principals; temporal stability, external validity, and cross‐context generalizability require further study.\n\n"]