Redd counts are commonly used to monitor the current population status, trends in abundance, and distribution of bull trout Salvelinus confluentus. In many cases redd counts are conducted at subjectively selected sites, and there has been limited evaluation of statistical sampling designs. We evaluated the utility of the generalized random tessellation stratified (GRTS) sampling design to determine bull trout population status through redd counts. We tested a sampling effort that would be economically practical to implement on a continuous basis in multiple drainages within the southeastern Washington and Oregon portions of the Columbia River plateau. We evaluated the logistics of a pilot application of the GRTS design, compared GRTS‐based estimates of redd abundance with those from census surveys, determined the precision of the GRTS estimates and the associated power for abundance comparisons, and compared the performance of the GRTS design with that of other probability sampling designs through simulation. A target of 50 sites per basin can be sampled by a two‐person survey crew multiple times over the spawning season. At that level of effort, the precision of redd abundance estimates ranges from 15% to 35%, depending on the patchiness of the redd distribution and the extent of the target population. These levels of precision are suitable for detecting a 30–70% change in redd abundance. Direct comparisons of GRTS‐based estimates with those obtained from a census showed mixed results. However, in a simulation study with three other probability sampling designs, GRTS consistently outperformed all but systematic sampling, which provided slightly better precision at intermediate sample sizes. Depending on the scale of inference, GRTS is useful in monitoring bull trout conservation units through redd counts, though a census may provide a more practical design for monitoring core areas as defined by the U.S. Fish and Wildlife Service.