Abstract Aim Coastal fishes have a fundamental role in marine ecosystem functioning and contributions to people, but face increasing threats due to climate change, habitat degradation and overexploitation. The extent to which human pressures are impacting coastal fish biodiversity in comparison with geographic and environmental factors at large spatial scale is still under scrutiny. Here, we took advantage of environmental DNA (eDNA) metabarcoding to investigate the relationship between fish biodiversity, including taxonomic and genetic components, and environmental but also socio‐economic factors. Location Tropical, temperate and polar coastal areas. Time period Present day. Major taxa studied Marine fishes. Methods We analysed fish eDNA in 263 stations (samples) in 68 sites distributed across polar, temperate and tropical regions. We modelled the effect of environmental, geographic and socio‐economic factors on α‐ and β‐diversity. We then computed the partial effect of each factor on several fish biodiversity components using taxonomic molecular units (MOTU) and genetic sequences. We also investigated the relationship between fish genetic α‐ and β‐diversity measured from our barcodes, and phylogenetic but also functional diversity. Results We show that fish eDNA MOTU and sequence α‐ and β‐diversity have the strongest correlation with environmental factors on coastal ecosystems worldwide. However, our models also reveal a negative correlation between biodiversity and human dependence on marine ecosystems. In areas with high dependence, diversity of all fish, cryptobenthic fish and large fish MOTUs declined steeply. Finally, we show that a sequence diversity index, accounting for genetic distance between pairs of MOTUs, within and between communities, is a reliable proxy of phylogenetic and functional diversity. Main conclusions Together, our results demonstrate that short eDNA sequences can be used to assess climate and direct human impacts on marine biodiversity at large scale in the Anthropocene and can further be extended to investigate biodiversity in its phylogenetic and functional dimensions.