One of the main challenges of modern engineering seismology is the mitigation of the adverse consequences of earthquakes. Although modern engineering techniques allow for the designing of earthquake‐resistant structures, the ability to predict more reliable ground‐motion estimates and associated uncertainties is needed. With this aim, this study investigated the possibility of using data recorded during an earthquake to improve the empirical ground‐motion prediction equations (GMPEs). In particular, we propose a procedure that updates the coefficients of an initial GMPE to account for the specific features of a seismic source and propagation medium. We applied the technique in the immediate postevent time period of three well‐recorded earthquakes that occurred recently in Italy and caused casualties and significant damage: the 6 April 2009 Mw 6.3 L’Aquila earthquake, the 20 May 2012 Mw 5.9 Emilia earthquake, and the 25 October 2012 Mw 5.2 Pollino earthquake. For possible future development, using the same earthquakes and the networks with which they were recorded, we also explored the potential of the technique as a possible real‐time application, as in the case of earthquake early warning systems.