A prototype of a generalized portrait of the activity of the Internet community for three major earthquakes that occurred in different regions of the Earth: Mexico, Japan and Afghanistan is discussed. The activity of the Internet community is a reaction to the second alarm system. A model of the reaction of Internet community activity to earthquake precursors is proposed. To diagnose "informational" harbingers of seismic danger, sets of query words semantically related to geophysical harbingers were used. The search for the repeatability of keywords was carried out using Internet search engines. The geographical binding of requests was taken into account. No effects of Internet community activity were detected in earthquake-prone regions. The analysis of information precursors of earthquakes was carried out according to the calculated spectra and amplitude diagrams of variations in the repeatability of the query words. Before the earthquakes in the earthquake-prone region, there was increased activity of the Internet community with statistically significant harbinger signals. The presented results are useful for use in artificial intelligence technologies.