Smaller grazers consistently show greater preference for recently burned patches than larger species. Energy optimization theory posits that this pattern is driven by small- versus large-bodied herbivores seeking to maximize energy intake by choosing high-quality recently burned grasses, or high-quantity unburned grasses, respectively. We propose that if burn preference is driven by an energy-maximization mechanism, then preference should change over time as grass regrows and progresses across the optimal feeding heights of herbivores of increasing body size. To test this, we used a camera trap array in the Serengeti National Park to quantify changes in the relative preference for burned patches of seven ruminant herbivore species. We compared observed patterns to simulation results from a grass production-herbivore patch selection model. Burn preference and herbivore body size scaled negatively for 6 months after fire, but this relationship disappeared after 7 months when smaller species stopped selecting burns, and larger herbivores selected burns after 10 months, in a reversal of classic grazer succession. Simulations recreated the former but not the latter relationship, suggesting that an energy-maximization mechanism can drive allometric scaling of burn preference immediately after fire, but over longer periods, grazer-driven feedbacks are required to explain large herbivore burn preferences.