AbstractTo solve spatial tasks, the human brain asks for support from the visual cortices. Nonetheless, representing spatial information is not fixed but depends on the reference frames in which the spatial inputs are involved. The present study investigates how the kind of spatial representations influences the recruitment of visual areas during multisensory spatial tasks. Our study tested participants in an electroencephalography experiment involving two audio–visual (AV) spatial tasks: a spatial bisection, in which participants estimated the relative position in space of an AV stimulus in relation to the position of two other stimuli, and a spatial localization, in which participants localized one AV stimulus in relation to themselves. Results revealed that spatial tasks specifically modulated the occipital event‐related potentials (ERPs) after the onset of the stimuli. We observed a greater contralateral early occipital component (50–90 ms) when participants solved the spatial bisection, and a more robust later occipital response (110–160 ms) when they processed the spatial localization. This observation suggests that different spatial representations elicited by multisensory stimuli are sustained by separate neurophysiological mechanisms.