Hovering means stationary flight at zero net forward speed, which can be achieved by animals through muscle powered flapping flight. Small bats capable of hovering typically do so with a downstroke in an inclined stroke plane, and with an aerodynamically active outer wing during the upstroke. The magnitude and time history of aerodynamic forces should be reflected by vorticity shed into the wake. We thus expect hovering bats to generate a characteristic wake, but this has until now never been studied. Here we trained nectar-feeding bats, Leptonycteris yerbabuenae, to hover at a feeder and using time-resolved stereoscopic particle image velocimetry in conjunction with high-speed kinematic analysis we show that hovering nectar-feeding bats produce a series of bilateral stacked vortex loops. Vortex visualizations suggest that the downstroke produces the majority of the weight support, but that the upstroke contributes positively to the lift production. However, the relative contributions from downstroke and upstroke could not be determined on the basis of the wake, because wake elements from down- and upstroke mix and interact. We also use a modified actuator disc model to estimate lift force, power and flap efficiency. Based on our quantitative wake-induced velocities, the model accounts for weight support well (108%). Estimates of aerodynamic efficiency suggest hovering flight is less efficient than forward flapping flight, while the overall energy conversion efficiency (mechanical power output/metabolic power) was estimated at 13%.