Dispersal emerges as a consequence of how an individual's phenotype interacts with the environment. Not all dispersing individuals have the same phenotype, and variation among individuals can generate complex variation in the distribution of dispersal distances and directions. While active locomotion performance is an obvious candidate for a dispersal phenotype, its effects on dispersal are difficult to measure or predict, especially in small organisms dispersing in wind or currents. Therefore, we analyzed the effects of larval swimming on dispersal and settlement of coral‐reef fish larvae using a high‐resolution biophysical model. The model is, to date, the only biophysical model of marine larval dispersal that has been statistically validated against genetic parentage estimates of larval origin and destination, and incorporates empirically‐estimated larval behaviors and their ontogeny. Larval swimming, in combination with depth, orientation and navigation behaviors, actually reduced dispersal distances compared to those of passive larvae. Swimming had no consistent effects on long distance dispersal, but increased the spread of settlement locations. Swimming speed, in contrast, did not consistently affect median dispersal distances, but faster swimming larvae had greater mean and maximum dispersal distances than slower swimming larvae. Finally, faster larval swimming speeds consistently increased the probability of settlement. Our analysis shows how larval swimming differentially affects multiple properties of dispersal kernels. In doing so, it indicates how selection could favor faster larval swimming to increase settlement and local retention, which may actually result in longer dispersal distances as a by‐product of larvae trying to locate habitat rather than to disperse greater distances. [ABSTRACT FROM AUTHOR]