Although assessment centers (ACs) are usually designed to measure stable competencies (i.e., dimensions), doubt about whether or not they reliably do so has endured for 70 years. Addressing this issue in a novel way, several published Generalizability (G) theory studies have sought to isolate the multiple sources of variance in AC ratings, including variance specifically concerned with competencies. Unlike previous research, these studies can provide a definitive answer to the AC construct validity issue. In this article, the historical context for the construct validity debate is set out, and the results of four large-scale G-theory studies of ACs are reviewed. It is concluded that these studies demonstrate, beyond reasonable doubt, that ACs do not reliably measure stable competencies, but instead measure general, and exercise-related, performance. The possibility that ACs measure unstable competencies is considered, and it is suggested that evidence that they do so may reflect an artefact of typical AC design rather than a “real” effect. For ethical, individual, and organizational reasons, it is argued that the use of ACs to measure competencies can no longer be justified and should be halted.