Artificial reefs (ARs) have been advocated and implemented as management tools for recreational fisheries, species conservation and habitat replacement. For ARs to function as substitute habitat for degraded natural reefs, they should perform as close as possible to local natural reefs, however this is seldom investigated. Here we evaluated the performance of new custom-designed reef structures (CDARs) as fish habitat. As a benchmark for their success, we compared fish abundance, diversity and community composition on CDARs to another commonly used AR type (Reef Balls (RBs)) and nearby natural reefs. Fish were monitored on all reef types over two recruitment seasons at three locations in Port Phillip Bay, Australia. Overall, there were no consistent differences in fish density among reef types, although densities on both AR designs were markedly lower than natural reefs at some locations. However, fish species richness on the CDARs was, on average, 2× higher than natural or RB reefs. There were large dissimilarities in fish community composition among reef types across all locations and years. These dissimilarities declined over time with the CDARs becoming more similar to natural communities than to RB reefs. Our results suggest that CDARs can play a role in reef fish conservation where natural reefs are under threat, supporting natural community structure and enhancing local biodiversity. Overall, our findings suggest that location of deployment, rather than design, has a more significant influence on fish abundances on ARs, whereas reef design is an important determinant of species diversity and community structure irrespective of location. ARs represent an important management tool for enhancing fisheries productivity and conservation in areas where reef habitat has been degraded or lost. However, failure to incorporate consideration of reef location and design into future AR deployments may lead to poor performance and failure to achieve restoration or conservation goals.