Observations of animals feeding in aggregations are often interpreted as events of social foraging, but it can be difficult to determine whether the animals arrived at the foraging sites after collective search [1–4] or whether they found the sites by following a leader [5, 6] or even independently, aggregating as an artifact of food availability [7, 8]. Distinguishing between these explanations is important, because functionally, they might have very different consequences. In the first case, the animals could benefit from the presence of conspecifics, whereas in the second and third, they often suffer from increased competition [3, 9–13]. Using novel miniature sensors, we recorded GPS tracks and audio of five species of bats, monitoring their movement and interactions with conspecifics, which could be inferred from the audio recordings. We examined the hypothesis that food distribution plays a key role in determining social foraging patterns [14–16]. Specifically, this hypothesis predicts that searching for an ephemeral resource (whose distribution in time or space is hard to predict) is more likely to favor social foraging [10, 13–15] than searching for a predictable resource. The movement and social interactions differed between bats foraging on ephemeral versus predictable resources. Ephemeral species changed foraging sites and showed large temporal variation nightly. They aggregated with conspecifics as was supported by playback experiments and computer simulations. In contrast, predictable species were never observed near conspecifics and showed high spatial fidelity to the same foraging sites over multiple nights. Our results suggest that resource (un)predictability influences the costs and benefits of social foraging. Using miniature tags, Egert-Berg et al. record bats’ movement and social interactions. Whereas species foraging on ephemeral resources search in groups, switching foraging sites, species foraging on predictable resources search alone, returning to the same sites. The results suggest a connection between resource predictability and group foraging.