AbstractEnvironmental DNA (eDNA) metabarcoding is a reliable method to assess taxonomic diversity, but the incompleteness of genetic reference databases prevents the assignation of many sequences to a given species. Functional diversity (FD) is a key biodiversity facet to monitor, but it requires the identification of all species within communities to account for their trait values. So, the ability of eDNA‐based inventories to estimate the “true” level of FD is unknown. Here, using fish surveys in two representative temperate and tropical rivers, with a quasi‐exhaustive genetic and trait database, we measured the bias and variability of different FD indices when calculated with uncertainty in taxonomic assignment at the genus and family level. Our results show that when measuring FD indices with species randomly chosen within genera and families, the bias cannot exceed 30% from real observed FD values. The variability is higher for species‐poorer communities and when those communities are composed of genera and families with high functional heterogeneity. Despite taxonomic uncertainty, our results demonstrate the potential of eDNA surveys to estimate reliable FD values.