Abstract Population densities of species have a predictable relationship with their body mass on a global scale. This relationship is known as the size–density relationship (SDR). The relationship was originally found to be directly opposite of metabolic rate scaling, which led to the hypothesis of energetic equivalence. However, recent studies have suggested that the SDR varies between clades. Specifically, the SDR for certain mammal clades has been found to be less negative than the relationship across all mammals. The aim of the present study is to estimate phylogenetic variation in the scaling relationship, using a data‐driven identification of natural phylogenetic substructure in the body size–density relation, and discuss its potential drivers. The classic model is often used to estimate natural population densities, and a further, practical aim is to improve it by incorporating variability among phylogenetic groups. We expand the model for the SDR relation of mammals to include clade‐specific variation. We used a dataset with population and body mass estimates of 924 terrestrial mammal species, covering 97 families, and applied an algorithm identifying group‐specific changes in the relationship across a family‐level phylogeny. We show increased performance in species density estimation is achieved by incorporating clade‐specific changes in the relationship compared to the classic model (increasing r2 from .56 to .74 and ΔAICc = 466). While the global SDR across clades was confirmed to be similar to previous findings (r = −.74), the relationship within all sub‐clades was less negative than the overall trend. Our results show that data‐driven identification of phylogenetic substructure in the size–density relation substantially improves predictive accuracy of the model. The less negative relationship within clades compared to the overall trend and compared to within clade metabolic scaling suggest that the energetic equivalence rule does not hold. This relationship shows that large species within clades use proportionally more energy than smaller species. Therefore, our results are consistent with a greater intra‐guild ecological impact of large‐bodied species via partial monopolisation of resources by the largest species of a given guild, and hence size‐asymmetric intra‐guild competition.