Rapid vegetation sampling methods based on visual estimation are useful for monitoring changes in rangeland vegetation composition because large spatial and temporal scales are often involved and have limited sampling resources available. Here we compared two sampling methods in their ability to detect changes in vegetation composition following rangeland development: 1) species percent cover estimates within subplots (the percent cover [PC] method) and 2) rankings of relative biomass of the 10 most abundant species across the whole plot and the ratio of two of them (the visual ranking [VR] method). Both methods were applied on 30 experimental plots at year 26 of a long-term factorial trial of five soil fertility levels and three sheep grazing intensities. Multivariate statistical methods showed significant effects of experimental treatments (fertilizer level and sheep grazing intensity) and of vegetation sampling method (VR vs. PC) on vegetation composition. Importantly, we detected no significant interactions involving sampling method, indicating that the effect of sampling method was consistent across experimental treatments. Effects of fertilizer on vegetation composition were an order of magnitude greater than the effect of sampling method, whereas the latter was twice as important as the effect of grazing. Results were robust to differential weights given to relative abundances vs. compositional changes. Differences between methods were primarily driven by the PC method giving lower abundance estimates of one species, lupin (a hybrid of Lupinus polyphyllus Lindl.), relative to the VR method. Our results support the use of the VR method as a rapid yet powerful method for monitoring changes in vegetation composition under rangeland development.