Environmental DNA metabarcoding is becoming a key tool for biodiversity monitoring over large geographical or taxonomic scales and for elusive taxa like soil organisms. Increasing sample sizes and interest in remote or extreme areas often require the preservation of soil samples and thus deviations from optimal standardized protocols. However, we still ignore the impact of different methods of soil sample preservation on the results of metabarcoding studies and there is no guidelines for best practices so far. Here, we assessed the impact of four methods of soil sample preservation commonly used in metabarcoding studies (preservation at room temperature for 6h, preservation at 4°C for three days, desiccation immediately after sampling and preservation for 21 days, and desiccation after 6h at room temperature and preservation for 21 days). For each preservation method, we benchmarked resulting estimates of taxon diversity and community composition of three different taxonomic groups (bacteria, fungi and eukaryotes) in three different habitats (forest, river bank and grassland) against results obtained under optimal conditions (i.e. extraction of eDNA right after sampling). Overall, the different preservation methods only marginally impaired results and only under certain conditions. When rare taxa were considered, we detected small but significant changes in MOTU richness of bacteria, fungi and eukaryotes across treatments, while the exclusion of rare taxa led to robust results across preservation methods. The differences in community structure among habitats were evident for all treatments, and the communities retrieved using the different preservation conditions were extremely similar. We propose guidelines on the selection of the optimal soil sample preservation conditions for metabarcoding studies, depending on the practical constraints, costs and ultimate research goals.