A method based on template matching is presented to detect and locate damage in buildings following severe shaking by an earthquake. The templates are constructed by finite‐element simulations of a suite of damage scenarios, with the solutions evaluated at the location (and orientation) of each sensor in the structure. The damage detection is carried out by cross‐correlating the templates with recordings acquired from earthquakes. A dense distributed network of sensors is important for detecting anomalies in the presence of ambient noise. The cross correlation of the templates with themselves provides a measure of the resolution of the damage location.