The importance of forests as privileged roosting and feeding areas for bats is well-established. How bats select their feeding places within large forests is less documented, while we may expect bats to react on habitat features at different spatial scales. In this study, ten teams recorded bat activity with bat detectors in 101 within-stand plots on two consecutive nights in a large French oak-dominated forest. Three groups of bat species that could be distinguished by their echolocation calls were considered. Splitting bat point counts into ten consecutive 1-minute periods made it possible to simultaneously model the probability of detection and the probability of occupancy. Explanatory variables included local stand features, landscape composition features and point count features. The probability of detection varied among teams and nights for the Nyctalus-Eptesicus group, among teams and with time of count and vegetation cover for the Pipistrellus group; none of these variables influenced the probability of detection of the Myotis group. This group showed a (moderate) preference for mature woods, while the Pipistrellus and Nyctalus-Eptesicus groups preferred regeneration stands (all groups within 250-m radius disks). The latter group was also more frequent at the edge of the forest than in the interior. Our study points to the fact that imperfect detectability needs to be accounted for in bat point count data analysis and that long rotations and natural regeneration are management practices beneficial to bat diversity in managed forests, even for forest-edge species.