We obtained 398 cytochrome c oxidase subunit I barcodes of 96 morphospecies of Lake Tanganyika (LT) cichlids from the littoral zone. The potential of DNA barcoding in these fishes was tested using both species identification and species delineation methods. The best match (BM) and best close match (BCM) methods were used to evaluate the overall identification success. For this, three libraries were analysed in which the specimens were categorized into Operational Taxonomic Units (OTU) in three alternative ways: (A) morphologically distinct, including undescribed, species, (B) valid species and (C) complexes of morphologically similar or closely related species. For libraries A, B and C, 73, 73 and 96% (BM) and 72, 70 and 94% (BCM) of the specimens were correctly identified. Additionally, the potential of two species delineation methods was tested. The General Mixed Yule Coalescent (GMYC) analysis suggested 70 hypothetical species, while the Automatic Barcode Gap Discovery (ABGD) method revealed 115 putative species. Although the ABGD method had a tendency to oversplit, it outperformed the GMYC analysis in retrieving the species. In most cases where ABGD suggested oversplitting, this was due to intraspecific geographical variation. The failure of the GMYC method to retrieve many species could be attributed to discrepancies between mitochondrial gene trees and the evolutionary histories of LT cichlid species. Littoral LT cichlids have complex evolutionary histories that include instances of hybridization, introgression and rapid speciation. Nevertheless, although the utility of DNA barcoding in identification is restricted to the level of complexes, it has potential for species discovery in cichlid radiations.