The risk of predation strongly affects mammalian population dynamics and community interactions. Bright moonlight is widely believed to increase predation risk for nocturnal mammals by increasing the ability of predators to detect prey, but the potential for moonlight to increase detection of predators and the foraging efficiency of prey has largely been ignored. Studies have reported highly variable responses to moonlight among species, calling into question the assumption that moonlight increases risk. Here, we conducted a quantitative meta‐analysis examining the effects of moonlight on the activity of 59 nocturnal mammal species to test the assumption that moonlight increases predation risk. We examined patterns of lunarphilia and lunarphobia across species in relation to factors such as trophic level, habitat cover preference and visual acuity. Across all species included in the meta‐analysis, moonlight suppressed activity. The magnitude of suppression was similar to the presence of a predator in experimental studies of foraging rodents (13·6% and 18·7% suppression, respectively). Contrary to the expectation that moonlight increases predation risk for all prey species, however, moonlight effects were not clearly related to trophic level and were better explained by phylogenetic relatedness, visual acuity and habitat cover. Moonlight increased the activity of prey species that use vision as their primary sensory system and suppressed the activity of species that primarily use other senses (e.g. olfaction, echolocation), and suppression was strongest in open habitat types. Strong taxonomic patterns underlay these relationships: moonlight tended to increase primate activity, whereas it tended to suppress the activity of rodents, lagomorphs, bats and carnivores. These results indicate that visual acuity and habitat cover jointly moderate the effect of moonlight on predation risk, whereas trophic position has little effect. While the net effect of moonlight appears to increase predation risk for most nocturnal mammals, our results highlight the importance of sensory systems and phylogenetic history in determining the level of risk.