Using high‐frequency identification, we provide evidence that Fed communication surprises have larger macroeconomic effects than surprise actions. Three ingredients are central to show this: structurally distinguishing between Fed actions and communication, controlling for the Fed information effect, and including the surprise measures directly in a vector autoregression (VAR) system instead of using them as instruments. We also compare the macroeconomic effects of Fed communication surprises relating to varying horizons into the future. Fed communication with a two‐year horizon appears most powerful during the effective lower‐bound period, consistent with theoretical predictions regarding Fed forward guidance.