Artificial Intelligence Literacy Divide: A New Threat to Sustaining Socioeconomic Disparities

Introduction: As Artificial Intelligence (AI) becomes a cornerstone of modern innovation, research increasingly focuses on AI literacy. Recognized as an advanced form of digital literacy, it encompasses competencies that enable individuals to understand, interact with, and critically evaluate AI systems (Lintner, 2024; Long & Magerko, 2020). While the traditional digital divide—i.e., unequal digital skills and access to broadband/hardware—is a known driver of economic inequality (Francis & Weller, 2022; Ochillo, 2022), there is growing concern that an AI literacy divide may further exacerbate these disparities. Existing research often relies on subjective self-assessments (Weber et al., 2023), potentially overlooking a “confidence-competence mismatch” where perceived knowledge misaligns with actual ability (Kruger & Dunning, 1999; Liu et al., 2025). This study addresses this gap by distinguishing between objective and subjective AI literacy. We examine how financial hardship is associated with these two dimensions among U.S. undergraduate students to inform policy interventions for equitable AI literacy.

Methods: Data were drawn from the 2024 Healthy Minds Study, a nationwide survey of U.S. university students. AI-specific modules were administered at two distinct institutions: a private Southern art school and a large public Midwestern doctoral university. We utilized two multiple linear regression models to analyze objective AI literacy (N = 608) and subjective literacy (N = 606). Objective literacy was measured using standardized scores based on the number of correct responses to six factual questions about AI in daily life; subjective literacy was measured using a three-point self-report scale. The primary predictor, financial hardship, was a four-category variable capturing both past and current status (no financial hardship, past only, current only, and both). The models controlled for AI use frequency, trust in generative AI, STEM major status, years in school, sex, race, and institutional type.

Findings: Results reveal a significant disconnect between actual and perceived AI literacy. Students experiencing both past and current financial hardship demonstrated significantly lower objective AI literacy (b = −.23, SE = .10, p < .05) than those with no history of hardship; however, no significant differences were observed across financial hardship groups in subjective literacy. Beyond economic status, objective literacy was higher among STEM majors and White and Asian students compared to their African American and Hispanic peers. Subjective literacy was associated with more frequent AI use and lower trust in generative AI. Additionally, male students and those at the private art school reported higher subjective literacy than female students and those at the public university.

Discussion: These findings suggest that the AI literacy divide is a “hidden” disparity. Because economically marginalized students may not perceive their literacy as lower despite having less objective knowledge, they may be less likely to seek out necessary training. This confidence-competence mismatch indicates that as AI becomes essential for the digital workforce, economic status continues to gate-keep the competencies required for socioeconomic mobility. Institutional efforts must move beyond addressing self-reported confidence and instead prioritize objective AI competencies, ensuring that AI serves as a pathway to opportunity rather than a mechanism for reinforcing existing disparities.