Paper The unequal adoption of ChatGPT exacerbates existing inequalities among workers

This study examines the adoption of ChatGPT among workers in occupations exposed to generative AI, using a large-scale survey linked to comprehensive administrative register data in Denmark. Drawing on responses from 18,000 workers across 11 exposed occupations, the research documents widespread uptake of the tool, particularly among younger and less experienced workers, while revealing significant inequalities in who is actually using it. A striking gender gap emerges: women are 16 percentage points less likely than men in the same occupation to have used ChatGPT for work. This disparity persists across occupations, adoption measures, and even among coworkers within the same workplace performing similar tasks. Additionally, despite evidence that generative AI tools offer the greatest productivity gains to workers with less prior expertise, suggesting potential to reduce existing workplace inequalities, users of ChatGPT were found to have earned somewhat more than non-users even before the tool's arrival. This pattern indicates that higher-achieving workers within their cohorts are disproportionately likely to adopt the technology. Workers broadly recognize the productivity potential of ChatGPT, estimating it could halve the time required for roughly a third of their job tasks. Yet perceived productivity gains are only weakly associated with actual use. Employer restrictions on tool usage and a perceived need for training emerge as the primary barriers to adoption. A randomized information experiment, in which workers were informed about potential time savings, produced no meaningful shift in usage behavior, confirming that these frictions causally prevent workers from capitalizing on the tool's benefits.

Get the Paper