Competitive interactions have been invoked as major drivers of ecological and evolutionary trends through time, but the strength and frequency of such interactions is notoriously difficult to evaluate. Here we use inferences from species co-occurrence patterns to determine the frequency of the extreme endmember of competition—competitive exclusion—in modern and fossil benthic marine invertebrate assemblages. Within environmentally and temporally well-constrained species pools, we examine the percentage of all species pairs that co-occur less often than expected by chance (segregated pairs) as a maximum measure of competitive exclusion. Segregated pairs are very rare, with most assemblages containing none, and the frequency of segregated pairs shows no trend throughout the Phanerozoic. Contrary to the competition-relatedness hypothesis, we find little evidence that segregated pairs are more common among closely related species except potentially for congeneric pairs in deeper water settings where physical disturbance is relatively low. Although taphonomy, post-mortem transport, and time-averaging could obscure original co-occurrence patterns, analyses suggest that these factors are unlikely to fully explain the low frequency of segregated pairs. Our findings support the hypothesis that predation and disturbance keep marine benthic communities below carrying capacity and thus reduce the potential for competitive interactions to dictate community composition in shallow waters.