摘要：Abstract Despite their ecological importance, sponges are often avoided in biodiversity studies and monitoring programs because they are notoriously difficult to identify using morphological or molecular methods. Here, we investigate the metabarcoding performance of universal degenerate cytochrome c oxidase subunit I (COI) primers in detecting species from this challenging phylum in a cryptobenthic community. Twenty‐two modified Autonomous Reef Monitoring Structures (ARMS) were deployed for 2 years in mesocosms receiving unfiltered seawater from an adjacent reef slope. Upon recovery, each unit was inspected by a marine sponge taxonomist who used a combination of taxonomy, imagery, and DNA barcoding (28S rRNA and COI) to identify sponges and generate a validated taxonomic richness value for each ARMS unit. A total of 69 unique sponge barcoded morphologies (BMs) were identified from the classes Calcarea, Demospongiae, and Homoscleromorpha. Metabarcoding identified 41 unique sponge molecular operational taxonomic units (MOTUs) from Demospongiae and Homoscleromorpha but the primers failed to amplify any species from the class Calcarea which comprised 22% of the BMs. Sponge richness did not differ between BMs and MOTUs assigned to the classes Demospongiae and Homoscleromorpha. However, assignments at the order and family level in Demospongiae underscore known limitations in sponge taxonomic resolution using the COI gene. The prevalence of false positives within the order Suberitida and the pervasiveness of false negatives within the order Haplosclerida highlighted both technical and biological constraints in the metabarcoding method. Overall, these results confirm the need for discretion in sponge MOTU assignments using universal degenerate barcoding primers that target a short fragment of the COI gene. However, they also demonstrate that COI metabarcoding is capable of capturing sponge richness from a complex natural community.
