Animal signalling systems outside the realm of human perception remain largely understudied. These systems consist of four main components: a signalling context, a voluntary signal, receiver responses and resulting fitness benefits to both the signaller and receiver(s). It is often most difficult to determine incidental cues from voluntary signals. One example is chemical disturbance cues released by aquatic prey during predator encounters that may serve to alert conspecifics of nearby risk and initiate tighter shoaling. We aimed to test whether disturbance cues are released incidentally (i.e. as a cue) or are produced voluntarily depending on a specific signalling context such as the audience surrounding the individual, and thus constitute a signal. We hypothesized that if receivers use disturbance cues to communicate risk among themselves, they would produce more (or more potent) disturbance cues when present in a group of conspecifics rather than when they are isolated (presence/absence of an audience) and use disturbance cues more when present alongside familiar rather than unfamiliar conspecifics (audience composition effect). We placed fathead minnows (Pimephales promelas) in groups with familiar fish, unfamiliar fish or as isolated individuals with no audience present, and then simulated a predator chase to evoke disturbance cues. We used bioassays with independent receivers to assess whether the disturbance cues produced differed depending on the signallers’ audience. We found evidence of voluntary signalling, as minnows responded to disturbance cues from groups of fish with tighter shoaling while disturbance cues from isolated minnows did not evoke a significant shoaling response (presence/absence audience effect). Receivers also increased shoaling, freezing and dashing more in response to disturbance cues from familiar groups compared to disturbance cues from unfamiliar groups or isolated minnows (audience composition effect). Together, these findings support our hypothesis that disturbance cues are used as an antipredator signal to initiate coordinated group defences among familiar conspecifics involving shoaling, freezing and dashing. This study represents the strongest evidence to date that chemicals released by aquatic prey upon disturbance by predators serve as voluntary signals rather than simply cues that prey have evolved to detect when assessing their risk of predation.