The impacts of climate change will depend on how human societies adapt to higher temperatures, but adaptation is very difficult to measure and typically not accounted for in estimates of the costs of climate change. Here, we study whether people living in warmer countries prefer higher temperatures for their recreational outdoor activities. To do so, we examine a unique dataset of mobile phone usage in parks and similar places covering two and a half years and more than 5000 locations across 130 countries. We first exploit quasi-random variation in weather from one day to the next to derive country-specific dose-response functions and then identify country-specific temperature levels at which outdoor activity peaks. In a second step, we regress these locally optimal temperatures on the annual mean climate. Our results point to substantial but only partial adaptation to average climatic conditions. For every degree Celsius increase in mean temperature, the optimal temperature for outdoor activity increases by about 0.45 degrees Celsius. Ignoring adaptation biases projections of future changes, which we illustrate with simulations.