When a uniformly illuminated surface is placed eccentrically on a dynamic textured background, after a few seconds, it is perceived to disappear and be replaced by the background texture. Such texture filling-in is thought to occur in retinotopic visual cortex, but it has proven difficult to distinguish the contributions of invisible target and visible background to signals measured in these areas. Here, we used magnetoencephalography to measure time-dependent brain responses in human observers experiencing texture completion. We measured responses specifically associated with the filled-in target, by isolating neural population signals entrained at the frequency of flicker of the target. When perceptual completion occurred, and the target became invisible, there was significant reduction in the magnetoencephalography power at the target frequency over contralateral posterior sensors. However, even a subjectively invisible target nevertheless evoked frequency-specific signals compared with a no-target baseline. These data represent evidence for a persistent target-specific representation even for stimuli rendered invisible because of perceptual filling-in.