The ‘abundant centre’ hypothesis states that species are more abundant at the centre of their range. However, several recent large‐scale studies have failed to find evidence for such a pattern. Here we used extensive global data of reef fishes to test the ‘abundant centre’ pattern, and to examine variation in the abundance patterns across species using life history and ecological traits. Marine habitat at a global extent: from Indo‐Pacific to Atlantic reefs. We used underwater visual estimates of fish abundance, containing 22,963 transects and 1,215 species. For each species we calculated the slope between abundance and distance to the range centre, with the range centre estimated using four different methods. We tested whether abundance patterns differ between the range core and margins using segmented regression. Meta‐analytic methods were used to synthesize results across species, and to test whether species traits can explain variation in the fit to the pattern among species. The method used to define the range centre had a large effect on the results. Nevertheless, in all cases we found large variation between species. Results of the segmented regression revealed that changes in abundance across the range core are very small and that steep declines in abundance happen only towards the range margins. Body size and mean abundance were the main traits affecting the fit to the pattern across species. We find large variation across species in the fit to the abundance centre pattern. Nevertheless, we do find support for a general pattern of a range core with high, but variable, abundance and steep decline in abundance towards the range periphery. Thus, species do tend to be rare at the range margins, making them sensitive to extirpation due to both natural and anthropogenic impacts.