Toward calibrated AI readiness in K–12 school-based practice: the roles of AI-related epistemic vigilance capacity and governance clarity

The responsible use of generative artificial intelligence (AI) in K–12 schools requires professionals to evaluate AI-generated information while maintaining ethical boundaries and accountability for decisions. This two-phase study examined calibrated AI readiness, defined as self-reported readiness to verify AI outputs, recognize ethical boundaries, and retain responsibility for AI-assisted professional decisions. In Phase 1, 425 school-based professionals completed a survey assessing AI-related epistemic vigilance capacity, autonomous regulatory motivation, governance clarity, and calibrated AI readiness. Structural equation modeling indicated that epistemic vigilance capacity ( β = .675, p < .001) and governance clarity ( β = .270, p < .001) were positively associated with calibrated AI readiness, whereas autonomous regulatory motivation was not a significant predictor ( β = .076, p = .156). An exploratory observed-score analysis found no significant interaction between epistemic vigilance capacity and autonomous regulatory motivation. In Phase 2, 25 school psychologists and school counselors participated in a one-month online pilot training program delivered through Moodle and Zoom. Self-reported calibrated AI readiness, measured using the same nine items as in Phase 1, increased descriptively from a pre-training mean of 4.20 to a post-training mean of 4.84 on a five-point scale. Scenario-based responses showed evidence of verification, privacy protection, contextual evaluation, and human accountability, although performance varied across specific indicators. Changes in self-reported readiness were not significantly associated with the continuous scenario-based reasoning index (Spearman's ρ = −.02, p = .906). Together, the findings suggest that perceived evaluative capacity and clear organizational guidance are important correlates of calibrated AI readiness, while also highlighting the value of combining self-report and applied reasoning assessments. Because Phase 1 was cross-sectional and Phase 2 lacked a comparison group, the results do not establish causal effects.