Current climate change alters the frequency and intensity of precipitation events, and the temporal autocorrelation between these events, i.e. their predictability. Little is known about how differences in predictability affect a plant’s chemical composition, and whether such differences may affect the provided ecosystem services. Here we experimentally tested this question by using a two-factorial field experiment, in which the forage legume Onobrychis viciifolia was exposed to more and less predictable daily (predictability of the summed simulated and natural daily precipitation) and seasonal precipitation over four consecutive years. Using an individual trait-based approach we show that less predictable precipitation led to higher aboveground and root biomass, and to higher nitrogen content in leaves and roots. Consequently, less predictable precipitation led to higher productivity. In contrast, lignin content and the C:N ratio were lower, which favours palatability, digestibility and litter decay. These results indicate that differences in precipitation predictability alter a plant’s chemical composition, plant-soil interactions, and that less predictable precipitation may improve ecosystem services provided by O. viciifolia, agricultural yields, and even, economic benefits, at least as long as no extreme events occur.