Many studies of wildlife poaching acknowledge the challenges of detecting poaching activities, but few address the issue. Data on poaching may be an inaccurate reflection of the true spatial distribution of events because of low detection rates. The deployment of conservation and law enforcement resources based on biased data could be ineffective or lead to unintended outcomes. Here, we present a rigorous method for estimating the probabilities of detecting poaching and for evaluating different patrol strategies. We illustrate the method with a case study in which imitation snares were set in a private nature reserve in South Africa. By using an experimental design with a known spatial distribution of imitation snares, we estimated the detection probability of the current patrol strategy used in the reserve and compared it to three alternative patrol strategies: spatially focused patrols, patrols with independent observers, and systematic search patterns. Although detection probabilities were generally low, the highest proportion of imitation snares was detected with systematic search strategies. Our study provides baseline data on the probability of detecting snares used for poaching, and presents a method that can be modified for use in other regions and for other types of wildlife poaching.