Land-use imposes an important potential threat on the aquatic ecosystems of riverine habitats. In this study, DNA metabarcoding was used to assess the effect of land-use on diatom assemblages, with a special focus on cropland area as an integrative proxy for several direct-acting pressures. The so-called taxonomy-free approach was tested using exact sequence variants (ESVs) without pre-clustering and without assigning them to traditional taxonomy. Our hypothesis is that the taxonomy-free approach gives better interpretation of the effect of agricultural land-use and it provides a more efficient index to indicate agricultural stress than the classical method with taxonomical clustering. Classical ordination techniques (PCA, NMDS) were performed to study the effects of cropland area proportion on diatom assemblages and a modified version of the Zelinka-Marvan equation for the index development. Results showed that (i) although ESVs provided better results when studying land-use effects on diatoms, taxonomic assignment after analysis was necessary to give ecological interpretations and that (ii) a better performing index could be developed by using the taxonomy-free approach. By using ESVs without taxonomic assignment, information on the ecology of sequences belonging to the same species and of unassigned sequences could be kept. New types of clustering methods are welcome in the future of biomonitoring where the delimitation of taxonomic units should be refined based on a higher emphasis on their ecology rather than on morphological or genetical criteria.