Marine Protected Areas (MPAs) offer a unique opportunity to test the assumption that fishing pressure affects some trophic groups more than others. Removal of larger predators through fishing is often suggested to have positive flow-on effects for some lower trophic groups, in which case protection from fishing should result in suppression of lower trophic groups as predator populations recover. We tested this by assessing differences in the trophic structure of reef fish communities associated with 79 MPAs and open-access sites worldwide, using a standardised quantitative dataset on reef fish community structure. The biomass of all major trophic groups (higher carnivores, benthic carnivores, planktivores and herbivores) was significantly greater (by 40% - 200%) in effective no-take MPAs relative to fished open-access areas. This effect was most pronounced for individuals in large size classes, but with no size class of any trophic group showing signs of depressed biomass in MPAs, as predicted from higher predator abundance. Thus, greater biomass in effective MPAs implies that exploitation on shallow rocky and coral reefs negatively affects biomass of all fish trophic groups and size classes. These direct effects of fishing on trophic structure appear stronger than any top down effects on lower trophic levels that would be imposed by intact predator populations. We propose that exploitation affects fish assemblages at all trophic levels, and that local ecosystem function is generally modified by fishing.