In recent years, both private and public organizations across contexts have begun implementing AI technologies in their recruitment processes. This transition is typically justified by improved efficiency as well as more objective, performance-based ranking, and inclusive selection of job candidates. However, this rapid development has also raised concerns that the use of these emerging technologies will instead increase discrimination or enhance the already existing inequality. In the present study, we first develop the concept of meta-algorithmic judgment to understand how recruiting managers may respond to automation of the hiring process. Second, we draw on this concept in the empirical assessment of the actual consequences of this type of transition by drawing on two large and unique datasets on employment records and job applications from one of Sweden's largest food retail companies. By comparing the outcomes of traditional and algorithmic job recruitment during this technological transition, we find that, contrary to the company's intentions, algorithmic recruitment decreases diversity. However, in contrast to what is often assumed, this is primarily not because the algorithms are biased, but because of what we identify as an unintended human–algorithmic interaction effect.