Plants host diverse microbial communities, but there is little consensus on how we sample and characterize these communities, and this has unknown consequences. Using root and leaf tissue from 20 showy milkweed (Asclepias speciosa) plants in the field, we compared two common sampling strategies by: 1) homogenizing after subsampling a small proportion of tissue (30 mg), and 2) homogenizing bulk tissue before subsampling 30 mg. Due to potential differences in richness and spatial distributions among microorganisms, we targeted bacteria, arbuscular mycorrhizal (AM) fungi and non-AM fungi in roots, as well as foliar fungal endophytes (FFE) in leaves. We also sampled FFE twice across the season using sampling strategy 1 to assess temporal dynamics, and we extracted DNA from all remaining homogenized bulk leaf tissues to determine the extent of potential undersampling. Bacterial richness was higher under sampling strategy 2, and all microbial groups except AM fungi differed in composition. Community overlap between the two sampling strategies increased when rare taxa were removed, but FFE and bacterial communities remained more different than alike and showed largely non-overlapping communities within individual plants. Increasing the extraction mass 10x also increased FFE richness ~10x, confirming the severe undersampling indicated in the sampling strategy comparisons. Even so, seasonal patterns in FFE communities were apparent, suggesting that strong drivers may be identified despite severe undersampling. Our findings highlight that current sampling practices poorly characterize many microbial groups and that increased sampling intensity is necessary to identify subtler patterns and to increase the reproducibility of studies.