Genotyping of maternally derived seed tissues from georefered seeds that moved away from their source tree yield direct estimates of seed dispersal distances when the location and the genotype of the fruiting tree are available. These estimates are instrumental in forecasting the response of plant communities to drivers of global change, such as fragmentation or the expansion of invasive species. Obtaining robust assessments of seed dispersal distances requires comparing reliable multilocus genotypes of maternally derived seed tissues and fruiting trees, as previously shown for angiosperm species. However, robust estimates of seed dispersal distances based on direct methods are rare in non-model gymnosperms due to the difficulty in isolating high quality DNA from inconspicuous maternally derived seed tissues. These tissues tend to yield low DNA quantities that increase the frequency of genotyping errors. Here, we deliver a step-by-step visual protocol used to identify and isolate different seed tissues of interest for dispersal studies: embryos (2n, bi-parentally derived), seed coats (2n, maternally derived), and megagametophytes (n, maternally derived). We also provide an optimised lab protocol used to obtain multilocus genotypes from the target seed tissue. These broadly applicable protocols proved successful both in avoiding contamination among different seed tissues and providing reliable multilocus genotypes.