Background The identification of Afrotropical hover flies is very difficult because of limited recent taxonomic revisions and the lack of comprehensive identification keys. In order to assist in their identification, and to improve the taxonomy of this group, we constructed a reference dataset of 480 COI barcodes of 83 of the more common nominal species from Ghana, Togo, Benin and Nigeria (W Africa), and evaluated its identification accuracy with three methods (K2P distance-based, Neighbor-Joining, and using SpeciesIdentifier). Results Results of the three methods were highly congruent and showed a high identification success. Nine species pairs showed a low (< 3%) mean interspecific K2P distance, with several cases of paraphyly and misidentifications. A high (> 3%) mean intraspecific K2P distance was observed in eight species, with several cases of para/polyphyly which may indicate the occurrence of cryptic species. Optimal K2P thresholds to differentiate intra- from interspecific K2P divergence were highly different among the three subfamilies (Eristalinae: 3.5%, Syrphinae: 5.8%, Microdontinae: 0.7-2%). Significance In addition to providing an alternative identification tool, our study indicates that DNA barcoding improves the taxonomy of Afrotropical hover flies by selecting (groups of) taxa that deserve further taxonomic study, and by attributing the unknown sex to species for which only one of the sexes is known. Yet, expanding the reference database by including less-common species with a poor taxonomic placement will prove challenging.