Background: Broad-scale monitoring of arthropods is often carried out with passive traps (e.g. Malaise traps) that can collect thousands of specimens per sample. The identification of individual specimens requires time and taxonomic expertise, limiting the geographical and temporal scale of research and monitoring studies. DNA metabarcoding of bulk-sample homogenates is faster and has been found to be efficient and reliable, but is destructive and prevents a posteriori validation of species occurrences and/or relative abundances. Non-destructive DNA metabarcoding from the collection medium has been applied in a limited number of studies, but further tests of efficiency are required in a broader range of circumstances to assess the consistency of the method. Methods: We quantified the detection rate of arthropod species when applying non-destructive DNA metabarcoding with a short (127-bp) fragment of mitochondrial COI on two types of passive traps and collection media: 1) water with monopropylene glycol (H2O-MPG) used in window-flight traps (WFT, 53 in total); 2) ethanol with monopropylene glycol (EtOH-MPG) used in Malaise traps (MT, 27 in total). We then compared our results with those obtained for the same samples using morphological identification (for WFTs) or destructive metabarcoding of bulk homogenate (for MTs). This comparison was applied as part of a larger study of arthropod species richness in silver fir (Abies alba) stands across a range of climate-induced tree dieback levels and forest management strategies. Results: Of the 53 H2O-MPG samples from WFTs, 16 produced no metabarcoding results, while the remaining 37 samples yielded 77 arthropod MOTUs in total. None of those MOTUs were shared species with the 389 morphological taxa (343 of which were Coleoptera) obtained from the same traps. Metabarcoding of 26 EtOH-MPG samples from MTs detected more arthropod MOTUs (233) and insect orders (11) than destructive metabarcoding of homogenate (146 MOTUs, 8 orders). Arachnida and Collembola were more diverse in EtOH-MPG samples, but Hymenoptera, Coleoptera and Lepidoptera were less represented than in homogenate. Overall, MOTU richness per trap similar for EtOH-MPG (21.81 MOTUs) than for homogenate (32.4 MOTUs). Arthropod communities from EtOH-MPG and homogenate metabarcoding were relatively distinct, with 162 MOTUs (53%) unique to the collection medium and only 71 MOTUs (23%) present in both treatments. Finally, collection medium did not reveal any significant changes in arthropod richness along a disturbance gradient in silver fir forests. We conclude that DNA metabarcoding of collection medium can be used to complement homogenate metabarcoding in inventories to favour the detection of soft-bodied arthropods like spiders.