AbstracteDNA metabarcoding has become a standard method for assessing wood‐inhabiting fungi and bacteria, yet determination of dead‐wood‐inhabiting beetles still relies on time‐consuming collection of beetle specimens. We thus tested whether beetle species can be identified by eDNA sequencing of wood in a mesocosm experiment that manipulated species assemblages. Dead wood samples were taken at exit holes of beetles and DNA was extracted and analyzed using two comparative methods: (i) metabarcoding with standard arthropod primers (421 bp) and (ii) using short species‐specific primers (120–264 bp) with Sanger sequencing. Results showed that beetle DNA was amplified by each of the two approaches, however, with (i) we detected only one non‐target saproxylic beetle species. In addition, we identified 80 different OTUs with four non‐targeted species of arthropods. For (ii) we detected the targeted species in two fresh beetle exit holes out of 20 samples. We suggest that, in contrast to fungi and bacteria, this eDNA metabarcoding approach is not able to reliably detect saproxylic beetles from wood samples, likely due to rapid degradation of their target DNA. Adapting such an approach for large‐scale analyses thus requires a better knowledge of degradation processes affecting DNA quality and quantity in wood.