Sentinel plasticine prey has been increasingly used to estimate predation pressure. The use of plasticine prey may, however, bias the results, as this method was originally designed to account for predation by organisms that can visually recognize the shapes and colors of their prey. To evaluate the limitations of using sentinel plasticine prey, we compared predator attack rates between real prey – dead and live mealworms, Tenebrio molitor L. (Coleoptera: Tenebrionidae) – and plasticine models in a monsoonal tropical rainforest of southeastern China. The attack rates by invertebrates were highest on dead prey followed by live prey and plasticine models, whereas the attack rates by vertebrates were lowest on dead prey, and did not differ between live prey and plasticine models. These results confirm that bias imposed by using the plasticine models is affected by the type of predators. In addition, we tested the validity and generality of the premise that predators can distinguish the shapes of plasticine model prey and preferentially attack a caterpillar‐like shape over other shapes. To test this hypothesis, we conducted three independent experiments in China, Papua New Guinea, and Finland. In the two latter localities, predation rates on plasticine caterpillars were higher than on models of other shapes, whereas in China, these differences were not significant. Taken together, our study suggests that plasticine models may underestimate the predation by invertebrates to a greater extent than predation by vertebrates, and the preference of model shape by predators may be locality‐specific, presumably due to differences in the composition of the predator community. We propose that predation be estimated on both live and plasticine prey in future studies to measure the potential bias imposed by using plasticine models and its variation among various habitats and predator groups.