The motion of GPS permanent stations during three earthquakes has been investigated with the use of Precise Point Positioning (PPP) technique and the seismological data. The study examines the ability of high-rate GPS observations to reflect the ground motion retrieved by the strong motion instruments (SM), considered to be more reliable and precise. The goal of this article is to show the sensitivity of GPS PPP kinematic high-rate positioning with position domain filtering using the band-pass Butterworth filter on small samples of position time series. The kinematic PPP approach in RTKLib software was used, supported by the CODE precise orbit and clock products to estimate positions from 5-hour long GPS phase datasets. Obtained position time series were reduced to 5-minute samples covering the time of co-seismic motion. The application of Butterworth band-pass filtering of GPS and seismological time series increased the agreement between them up to 72 % in terms of correlation, resulting in correlations within the range 0.34 to 0.99. The comparison of peak ground displacements (PGD) revealed that for Italian events, GPS–SM absolute value of the average difference is 6 mm with GPS–SM distances within the range of 0.05 to 2.14 km. In all analysed earthquakes, the agreement between GPSgrams and seismograms in terms of the first P-arrival polarity was checked and it was found that it is consistent in all cases. This confirms the GNSS technique capability for determining fault plane solution for earthquakes with magnitudes over 6.