Telemetry is increasingly being used to estimate population-level survival rates. However, these estimates may be affected by the detectability of telemetry tags and are reliant on the assumption that telemetry data represent the movements of the tagged fish. Predation on tagged fish has the potential to bias survival estimates, and unlike the issue of detectability, methods to correct for the resulting bias (termed “predation bias”) are not yet developed. In an acoustic telemetry study on inner Bay of Fundy Atlantic salmon (Salmo salar) smolts during 2008 and 2011, unusual tag detection patterns were indicative that some data may have been representative of the movements of predators rather than smolts. To incorporate predation effects into the resulting survival estimates, a suite of 11 summary migration metrics were compared between Atlantic salmon smolts and striped bass (Morone saxatilis). Cluster analyses revealed that 2.4% to 13.6% of tags implanted in smolts exhibited migration patterns more similar to striped bass than to other smolts, which was interpreted here as evidence of predation. Reassigning the fate of these tags as “depredated–died” reduced estimated survival from 43.5% to 41.1% in 2008 and from 32.6% to 19.0% in 2011 relative to a traditional mark–recapture model, illustrating the effect of predation bias in this case study.