Inclusive generative AI and the dynamics of inequality: A capability-resource model of adoption disparities

The rapid diffusion of generative artificial intelligence (GenAI) has renewed concerns that emerging technologies may deepen rather than reduce digital inequality. Using nationally representative survey data from South Korean adults aged 19–69 who use the internet daily (N = 4344), this study examines how socioeconomic status (SES) shapes GenAI adoption. Lower-SES individuals—defined by the intersection of low income and limited education—exhibit a substantial adoption gap (4.8% vs. 13.3%). Across the twelve items, SES differences were most consistent for capability-related barriers (knowledge deficits and perceived complexity). Resource-related differences were limited, with only cost burden showing a small gap, while perception-based items exhibited weak and non-systematic SES gradients. Digital capability is positively associated with adoption but does not significantly moderate the SES gap, supporting an additive pattern in which socioeconomic disadvantage and capability deficits operate as parallel constraints. Drawing on digital divide theory and a theory-informed barrier typology, we locate where SES gradients concentrate (capability vs. resource vs. perception) and test whether capability attenuates SES disparities. These findings suggest that GenAI's democratizing potential may be conditional rather than automatic, particularly because the technology amplifies cognitive productivity when used effectively. Because GenAI participates in content generation, problem solving, and decision support, early inequalities in access may translate into unequal opportunities to accumulate collaborative AI experience over time. Without coordinated efforts addressing structural and capability-based barriers, AI diffusion may reproduce existing socioeconomic inequalities.