Despite the increasing popularity of DNA metabarcoding in the assessment of aquatic ecosystems using fish eDNA or ichthyoplankton, challenges have hampered its broader application in the Neotropical freshwaters. Using five mock communities composed of fish species from two Neotropical river basins, we evaluated the influence of DNA concentration and choice of mitochondrial 12S molecular markers (MiFish, NeoFish and Teleo) on species detection and Relative Read Abundance (RRA) using DNA metabarcoding. Of the three 12S markers analysed, only MiFish detected all species from all mock communities. The performance of a taxonomy-free approach using ASV/MOTUs was not as precise as assigning DNA reads to species using a curated 12S library that includes approximately 100 fish species, since more than one ASV/MOTU was observed for the same specimen. Thus, here we showcase the importance of a custom reference database to allow precise assignment of Neotropical fish species in metabarcoding studies and that the RRA is dependent on community composition, marker and DNA concentration. We highlight the importance of controlled experiments using known species communities before large investments are made in assessing biodiversity using non-invasive methods that apply DNA metabarcoding.