Abstract Quantitative forecasts have become increasingly prominent as tools for aiding public understanding of sociopolitical trends. But how much, and what, do people learn from quantitative forecasts? In this note, we show through a preregistered survey experiment that real forecasts of the 2022 French presidential election significantly affect expectations of the election result. The direction of that effect hinges on how the forecast is presented. Voters become more accurate and precise in their predictions of each candidate’s vote share when given forecast information in the form of projected vote share. Forecasts presented as numerical probabilities make such expectations less accurate and less precise. When combined, the effects of both forms on vote share expectations tend to cancel out, but jointly boost voters’ ability to identify likely winners. Our findings have implications for the public communication of quantitative information.