Ecological restoration - the rebuilding of damaged or destroyed ecosystems - is a critical component of conservation efforts, but is hindered by inconsistent, unpredictable outcomes. We investigated a source of this variation that is anecdotally suggested by practitioners, but for which empirical evidence is rare: the weather conditions during the first growing season after planting. The idea of whether natural communities face long-term consequences from conditions even many years in the past, called historical contingency, is a debated idea in ecological research. Using a large dataset (83 sites) across a wide geographic distribution (three states), we find evidence that precipitation and temperatures in the planting year (2-19 years before present) affected the relative dominance of the sown (native target species) and non-sown (mostly non-native) species. We find strong support for lasting planting year weather effects in restored tallgrass prairies, thereby supporting the historically contingent model of community assembly in a real-world setting.