Public Trust in AI-Driven Science: The Moderating Role of Self-Reported AI Fluency in the European Union

The increasing integration of Artificial Intelligence (AI) into scientific discovery poses a fundamental challenge to public trust and established sources of epistemic authority. While traditional formal education has long been considered a predictor of confidence in science, this study investigates whether a citizen’s self-reported AI/digital fluency moderates this relationship, focusing on trust in AI-generated scientific discoveries (qa7_1). Employing a Multilevel Moderated Regression analysis on cross-sectional Eurobarometer data (N = 26,404) from 28 EU countries, we tested the hypothesis that fluency serves as an enabling condition for educational capital to translate into acceptance. The data reveal a strong positive main effect of AI Fluency (β = 0.373, p < .001) on trust. Critically, the main effect of Formal Education was not statistically significant (β ≈ −0.000, p = 0.554,95% CI [−0.001,0.000]), indicating no detectable linear relationship in this model after accounting for AI fluency. A statistically significant but practically minimal positive interaction (β = 0.001, p = 0.038,95% CI [0.00006,0.001]) was observed. The model accounted for 24% of variance in trust (R2 = 0.24, p < .001). These findings suggest that AI Fluency is a substantially stronger predictor of public trust than formal education in the context of AI-driven science. If this correlational pattern reflects causal mechanisms, policy efforts to build trust should prioritize practical digital literacy and technological self-efficacy alongside general education initiatives.