Presence/absence sampling designs are commonly used approaches for documenting the occurrence of benthic invertebrates, but these methods often provide biased indices of occupancy or abundance. Unionids are particularly susceptible to bias associated with nondetection because of their generally low abundance and burrowing behavior. We present a case study in which we used an occupancy-modeling approach that incorporates estimation of detection probability to assess the occurrence of unionids at 39 locations along a 119-km reach of Flint River, Georgia. We fitted models relating site- and reach-level variables to detection and occupancy for 3 species of conservation concern and contrasted inferences drawn between models accounting for variable imperfect detection and models assuming constant imperfect detection. We collected 20 unionid species among 39 sites. Among species, detection probabilities averaged 0.25 and ranged from 0.01 to 0.69, and estimated occupancy averaged 0.56 and ranged from 0.03 to 1.00. Estimated occupancy across all species was, on average, 25.78% greater than naïve occupancy (i.e., proportion of sites without accounting for incomplete detection). Detection probabilities varied by species, time, macrohabitat, and searcher experience. Inferences regarding factors affecting species occupancy differed among models contrasting detection as variable or constant but imperfect. The occupancy-modeling approach was a useful and efficient method to assess the distribution and habitat relationships of unionids in the lower Flint River. We think that detection probability is an underused but necessary parameter that should be evaluated during sampling to provide a measure of confidence in benthic invertebrate data and to account for potential sources of bias.