AI Literacy and Critical Information Evaluation: Assessing the Role of Academic Librarians in Mediating Student Verification of Synthetic and Hallucinated Scholarly Content

Abstract

Background: The proliferation of generative artificial intelligence across higher education has heightened the risk of algorithmic hallucinations and phantom citations contaminating student research. Despite the severity of these epistemic risks to scholarly integrity, empirical assessments evaluating how academic librarians actively mediate student verification of synthetic scholarly content remain limited.

Objective: Anchored in frameworks of cognitive risk, AI-mediated information literacy, and epistemic authority, this study evaluated the reported prevalence of phantom citations in reference consultations, assessed academic librarians' self-perceived AI literacy and mediation practices, and tested the predictive and moderating relationships governing student verification adoption.

Methodology: A cross-sectional survey design sampled 246 academic librarians across federal, state, and private universities. Data gathered via a validated questionnaire were analysed using descriptive statistics, Pearson product-moment correlation, standard multiple linear regression, and hierarchical moderated regression with simple slopes probing.

Result: Librarians reported a high prevalence of student encounters with synthetic literature (Mean = 3.73), dominated by citations pairing authentic authors with fabricated titles. Technical AI literacy significantly correlated with proactive verification mediation (r = 0.584, p < 0.001). Instructional guidance intensity strongly predicted students' adoption of lateral verification strategies (Beta = 0.592, p < 0.001), explaining 37.5% of the variance. Furthermore, perceived cognitive authority significantly moderated this relationship (B = 0.174, p = 0.002), more than doubling the positive impact of instructional frequency on phantom citation detection under high authority conditions.

Conclusion: Academic librarians serve as vital institutional safeguards against generative hallucinations. Enhancing librarians' technical AI literacy, instructional intensity, and cognitive authority is essential for instilling robust lateral verification habits and preserving scholarly integrity.