Rapid identification of felt earthquakes is essential for determining public earthquake information. We present a new method to detect such earthquakes that hinges on the ubiquity of smartphones and the accurate geolocation that they offer. More precisely, the method is based on launches by its users of the LastQuake app, the European‐Mediterranean Seismological Centre (EMSC) app providing rapid global earthquake information. Similar to two other existing methods, one based on the analysis of earthquake information website traffic and the other based on the publication on Twitter of earthquake related messages, it exploits the online reaction of eyewitnesses following ground shaking. Its time performance is shown to depend on the number of app users in the epicentral region and whether the earthquake happens during day or night. Over the 16‐month study period, the observed time difference between the arrival of the P waves and the app launch times is typically 10 s longer at night than during the day. These reaction times can significantly decrease during a sequence of earthquakes affecting the same region, leading in the best cases to earthquake detection times as fast as 20 s from earthquake origin time. Eyewitnesses’ locations determined from app launches also map the felt area. In turn, in some cases, the surface of the felt area could offer a first‐order magnitude estimate within a few tens of seconds of their occurrence for small‐magnitude earthquakes and in a few minutes for larger ones. The analysis of online reaction of eyewitnesses not only offers seismological information that complements that derived from seismological networks (eg, rapid identification of felt earthquakes, mapping of the felt area) but also provide insights into eyewitnesses’ behaviors and expectations during and immediately after a tremor.