ARGOS tracking data are frequently used to infer feeding locations of marine predators. Most track-based methods rely on the assumption that feeding takes place within regions of area-restricted search (ARS); however, it is unclear whether the spatial accuracy and temporal resolution of ARGOS-quality data are sufficient to extract the location of ARSs at the resolution of individual feeding bouts. Using ARGOS tracking data from northern elephant seals (Mirounga angustirostris) and Laysan albatrosses (Phoebastria immutabilis), we tested several track-based estimators of feeding locations against independent feeding proxies (dive type or shape and number of landing events). The track-based methods were: turn angle, transit rate, first-passage time, and fractal dimension. None of these methods provided a reliable estimate of putative feeding activity at the species level, as measured by dive-type analysis or landing events. At the individual level, variation in agreement between track-based estimates and feeding proxies highlighted the importance of considering the effect of spatial scale. Biological justification of track-based metrics should be more carefully assessed before assigning particular functions to individual tracks (feeding/foraging/searching/transit) or groups of tracks (biological hotspots).