Recently available data from online job vacancies have enabled analyses that move beyond across-occupation variation to also include within-occupation variation in workers’ task-specific skills. However, analyses of job vacancy data are limited by the fact that information on the hired worker(s) is hidden. To overcome this issue, I develop a novel, pseudo-individual match between Danish job vacancy data and register data. With data on the hired worker(s) for each online job vacancy, I can test how the employment of skills and the returns to skills depend on the gender of the worker. I use the matched employer-employee-vacancy data to show that women face significantly lower returns to cognitive, character, customer service, financial, and specific computer skills when compared to men while controlling for both occupation and firm fixed effects.