Eelgrass (Zostera marina) populations occupying coastal waters of Alaska are separated by a peninsula and island archipelago into two Large Marine Ecosystems (LMEs). From populations in both LMEs, we characterize genetic diversity, population structure, and polarity in gene flow using nuclear microsatellite fragment and chloroplast and nuclear sequence data. An inverse relationship between genetic diversity and latitude was observed (heterozygosity: R² = 0.738, P < 0.001; allelic richness: R² = 0.327, P = 0.047), as was significant genetic partitioning across most sampling sites (θ = 0.302, P < 0.0001). Variance in allele frequency was significantly partitioned by region only in cases when a population geographically in the Gulf of Alaska LME (Kinzarof Lagoon) was instead included with populations in the Eastern Bering Sea LME (θ_p = 0.128–0.172; P < 0.003), suggesting gene flow between the two LMEs in this region. Gene flow among locales was rarely symmetrical, with notable exceptions generally following net coastal ocean current direction. Genetic data failed to support recent proposals that multiple Zostera species (i.e. Z. japonica and Z. angustifolia) are codistributed with Z. marina in Alaska. Comparative analyses also failed to support the hypothesis that eelgrass populations in the North Atlantic derived from eelgrass retained in northeastern Pacific Last Glacial Maximum refugia. These data suggest northeastern Pacific populations are derived from populations expanding northward from temperate populations following climate amelioration at the terminus of the last Pleistocene glaciation.