Freshwater biomonitoring under the EU Water Framework Directive (WFD) relies on multiple biological quality elements, with macrozoobenthos being one of the most widely used, although still constrained by time-consuming morphological identification and the continued decline in taxonomic expertise. These constraints, together with ongoing rapid biodiversity loss, create an urgent need for high-resolution, scalable assessment methods such as DNA metabarcoding. Here, we evaluated this molecular approach as a WFD-compatible method and compared the performance of bulk and environmental DNA (eDNA) sampling for ecological quality classification in Slovakia. During spring 2022–2023, we sampled 34 WFD long-term monitoring sites using (i) bulk macroinvertebrate kick-samples, and (ii) water and (iii) sediment eDNA. A ~420 bp fragment of the mitochondrial COI gene was amplified using the BF3/BR2 primers, and sequenced. Filtered reads were clustered into OTUs based on 97% identity and taxonomically assigned using the BOLD database. Raw read counts were used as abundance proxies to calculate 12 ASTERICS metrics and multimetric Ecological Quality Ratios (EQRs). Sequencing yielded > 32.1 million reads and 1,205 target OTUs, of which 702 corresponded to 558 Linnaean species. Bulk samples captured the highest richness, followed by water and then sediment eDNA. Bulk samples classified 22 sites as very good Ecological Quality Class (EQC), whereas water and sediment samples frequently reduced EQC by two to three classes. EQC differences were driven mainly by presence/absence metrics (EPT taxa and BMWP), while several abundance-dependent metrics remained stable. Bulk metabarcoding provides a comprehensive signal but still inflates presence/absence metrics, highlighting the need for recalibration and optimised eDNA sampling, including primer selection, to enable comparable non-invasive assessments.