Previous research has suggested that videos can be used to experimentally manipulate social stimuli. In the present study, we used the California scrub-jays' cache protection strategies to assess whether video playback can be used to simulate conspecifics in a social context. In both the lab and the field, scrub-jays are known to exhibit a range of behaviours to protect their caches from potential pilferage by a conspecific, for example by hiding food in locations out of the observer's view or by re-caching previously made caches once the observer has left. Here, we presented scrub-jays with videos of a conspecific observer as well as two non-social conditions during a caching period and assessed whether they would cache out of the observer's "view" (Experiment 1) or would re-cache their caches once the observer was no longer present (Experiment 2). In contrast to previous studies using live observers, the scrub-jays' caching and re-caching behaviour was not influenced by whether the observer was present or absent. These findings suggest that there might be limitations in using video playback of social agents to mimic real-life situations when investigating corvid decision making.