Seagrass beds are an important habitat for multiple species of fish, invertebrates, algae, and waterfowl. The decline of seagrass populations worldwide has spurred efforts to restore seagrasses to suitable habitat. The major focus has been to simply increase the number of plants in new and existing seagrass beds, rarely considering the conservation of genetic diversity from the very outset of a restoration project. Here we consider genetic diversity in a mesocosm-based test of a restoration method that uses seeds gathered from three San Francisco Bay populations of Zostera marina with different degrees of sexual vs. clonal reproduction. Eighteen mesocosms were sowed with seeds from one of the three populations. DNA was extracted from 128 seedlings and scored at six microsatellite loci. As expected, seeds from the annual population resulted in new plants that harbored more genetic diversity (H_O = 0.339) than plants grown from the two perennial populations' seeds (H_O = 0.140 and 0.076). In addition, the plants from the annual population were genetically differentiated from the other two (pairwise F_ST = 0.142 – 0.143, p < 0.0001), which were not differentiated from each other (pairwise F_ST = 0.018, p = 0.090). Shoot densities were highest initially in the mesocosms from the perennial populations, but by the end of the experiment the mesocosms containing annual-derived plants had the highest shoot density. We also compared patterns of genetic diversity in the mesocosms to genetic diversity and differentiation in the field source populations. Overall, these results suggest that the amount of genetic diversity present in seed donor populations may be an important predictor of long-term persistence of restored seagrass meadows. We recommend additional field studies to explicitly test this pattern of slower initial growth in annually-seeded areas, followed by density increases surpassing those in perennially-seeded areas.