Invertebrates inhabiting grasslands benefit from uncut grass refuges, yet effects on the plant community have not been properly quantified. We experimentally investigated the effects on the vegetation of two different types of refuges. While both consisted in not mowing 10–20% of a meadow area, they differed in their rotation frequency: (1) in within-year rotational refuges (WYRR), the location of the refuge within a meadow was changed at each mowing operation, usually twice a year; (2) in between-years rotational refuges (BYRR), the refuge changed location only between years. A third mowing regime without any refuge was included as control (C) for comparison. The study was conducted in thirty extensively managed meadows across the Swiss lowlands. The vegetation was sampled at two 1-m2 plots within each of the four strata defined by a stratified random design that accounted for the spatial location of the uncut refuge over the years. There were no overall significant negative effects of WYRR on plant species richness and composition at the meadow scale, although a small negative effect was detected locally (i.e. at the refuge scale) where a WYRR had been implemented more than once in the preceding three years. Leaving BYRR negatively impacted plant species richness (-11%), even reducing the number of indicator plant species by 22% (from 4.5 to 3.5 per 2 m2), regardless of when and where refuges were left uncut. A beta-diversity analysis revealed no difference at community level between the two refuge types and control meadows. Previous studies had evidenced positive effects of uncut refuges on herbivore and pollinator communities, while this study shows that the plant community is not affected as long as the location of the refuge is changed at each mowing operation. We thus recommend this measure for promoting biodiversity in extensively managed grasslands.