Abstract The link between dispersal traits and patterns of community assembly remains a frontier in understanding how vertebrate communities persist in fragmented landscapes. Using experimental release trials and intensive field surveys of bird communities in fragmented forests of the Peruvian and Colombian Andes, we demonstrate that morphological traits related to movement (1) predict experimental flight performance and (2) exhibit dispersal‐mediated environmental filtering at the community scale. After correcting for body size, four traits hypothesised to influence flight ability (wing length, wing pointedness, wing loading, and eye size) predicted distance flown across a hostile experimental landscape, with successful species having significantly longer pointed wings, carrying less mass per unit wing area (i.e. lower wing loading), and having smaller eyes. Species with larger eyes also displayed increased flight latency, potentially due to disability glare. At the community scale, we detected a gradient of dispersal‐mediated environmental filtering in fragments compared to reference forest within the same landscape, with relative differences in trait values explained by the temporal and spatial extent of patch isolation. In the Colombian landscape where fragments had been isolated for >60 years, communities were filtered for species with long and narrow wings and small eyes, especially within the most spatially isolated fragments. We observed the opposite pattern in the more recently fragmented Peruvian landscape (15–30 years): communities within fragments tended to have shorter and more rounded wings compared to those in nearby contiguous forests, suggesting that dispersal‐limited species accumulate in the initial years following patch isolation due to “restricted dispersal” and represent an extinction debt yet to be paid. Our results (1) experimentally validate the use of morphological traits as proxies for movement ability in fragmented landscapes, (2) demonstrate that visual acuity functions as a novel dimension of dispersal limitation and (3) quantify how the spatial and temporal components of patch isolation produce a gradient in dispersal‐mediated environmental filtering and extinction debt for communities inhabiting fragments. Read the free Plain Language Summary for this article on the Journal blog.