Beyond the Connection: AI Literacy As a Second-Level Digital Divide In an Urban Media Ecology

Background and Purpose: The rapid integration of Artificial Intelligence (AI) represents a historic technological shift, necessitating specialized competencies for equitable societal participation. Grounded in a social equity framework, this study conceptualizes AI literacy as a manifestation of the second-level digital divide, moving beyond mere physical access to focus on disparities in skills and usage. While traditional digital divides are well-documented, less is known about how digital immersion and social stratification shape AI-specific competencies in diverse urban environments. This research examines the structural factors associated with AI literacy among residents of San Antonio, Texas, to inform interventions that prevent emerging technologies from reinforcing historical socioeconomic disparities.

Methods: Using a cross-sectional survey, data were collected from 647 adult residents. Recruitment focused on public library branches to ensure broad geographic and socioeconomic coverage. Exploratory Factor Analysis (EFA) confirmed a two-factor structure for AI literacy: Perception (critical awareness of AI’s societal roles and ethical implications) and Utilization (practical operational skills and evaluative competence). Multiple linear regression was employed to identify predictors of total AI literacy and its sub-dimensions, controlling for digital immersion (daily internet hours and device variety), anxiety, and social-structural factors such as education, age, and ethnicity.

Results: Regression models significantly explained the variance in AI literacy (R² = .061, p < .0001). Digital immersion emerged as the primary driver of practical Utilization skills; every unit increase in internet usage predicted a significant rise in functional competence (B = 1.03, p < .0001). In contrast, formal education was more strongly associated with critical Perception (p < .001). A notable "digital native myth" was identified: older age predicted higher Perception of AI's importance (B = 0.04, p < .001) but was unrelated to actual Utilization, suggesting that for older adults, age is linked to symbolic valuation rather than functional proficiency. Furthermore, self-employed individuals reported significantly higher AI anxiety (p = .040), reflecting economic vulnerability to AI-driven disruption.

Conclusions and Implications: These findings suggest that closing the AI divide is not merely a technical problem but requires addressing foundational inequalities. The results support a dual-track intervention model: expanding opportunities for hands-on digital immersion to build operational skills while strengthening formal educational foundations to foster critical appraisal. For social work practice, this necessitates culturally and linguistically responsive programs, particularly for Spanish-dominant residents who are currently underrepresented to ensure that AI serves as a tool for empowerment rather than a mechanism for further marginalization.