Many terrestrial endotherm food webs constitute three trophic level cascades. Others have two trophic level dynamics (food limited herbivores; plants adapted to tackle intense herbivory) or one trophic level dynamic (herbivorous endotherms absent, thus plants compete for the few places where they can survive and grow). According to the Exploitation Ecosystems Hypothesis (EEH), these contrasting dynamics are consequences of differences in primary productivity. The productivity thresholds for changing food web dynamics were assumed to be global constants. We challenged this assumption and found that several model parameters are sensitive to the contrast between persistently warm and seasonally cold climates. In persistently warm environments, three trophic level dynamics can be expected to prevail almost everywhere, save the most extreme deserts. We revised EEH accordingly and tested it by compiling direct evidence of three and two trophic level dynamics and by studying the global distribution of felids. In seasonally cold environments, we found evidence for three trophic level dynamics only in productive ecosystems, while evidence for two trophic level dynamics appeared in ecosystems with low primary productivity. In persistently warm environments, we found evidence for three trophic level dynamics in all types of ecosystems. The distribution of felids corroborated these results. The empirical evidence thus indicates that two trophic level dynamics, as defined by EEH, are restricted to seasonally cold biomes with low primary productivity, such as the artic–alpine tundra and the temperate steppe.