We derived a model to predict site selection by drifting prey in streams. This model considers the conflicting demands between feeding and avoiding both benthic and drift predators. Our analysis suggests a ranking of site qualities based on the ratio of food acquisition rate to benthic predation risk (termed site value). Drifting organisms should accept a given site type when its site value exceeds the expected value of drifting, which is the average site value prey are likely to encounter adjusted for the costs of drifting. These costs are represented as drift predation risk and time that is lost from foraging as prey search for a better site. We contend that prey should rank site types in descending order based on site value and sequentially add site types to the acceptable category in order of highest rank until the addition of the next lowest ranked site decreases rather than increases the expected value of drifting. The best sites should be those with high food availability and low benthic predation risk. Prey should reject a site when drift predation risk is low, the proportion of acceptable site types is high, and the rate that prey settle from the drift is high. Interestingly, prey with the greatest locomotory ability should have the greatest propensity to drift because these individuals spend less time drifting, making them less susceptible to drift predators. Our model provides a framework to simultaneously integrate the effects of food availability, benthic predation risk, and drift predation risk on site selection of actively drifting prey.