We investigate the ability of GPS to detect volcanic plumes at Mt. Etna, Italy. We use a robust statistical approach to highlight whether the presence of a volcanic plume in the atmosphere may really affect the GPS undifferenced post-fit phase residuals. The proposed method has been tested for the September 4–5, 2007 activity of Mt. Etna. This eruption produced powerful lava fountains forming a weak, a few kilometers high plume for several hours, representing typical activity at Etna over the last 5 years. We analyzed data from nineteen Etna permanent GPS stations located on the volcano flanks at different heights and applied a statistical test based on four main steps: (a) realization of a simplified model representing the volcanic plume in atmosphere; (b) evaluation of the GPS satellite and station couples intersecting the plume; (c) calculation of the volcanic plume region crossed by the GPS signal; (d) application of a robust statistical test in order to see whether the volcanic plume affected the GPS signals. Results show that during the September 4–5, 2007 explosive activity, the GPS residuals definitely include the contribution of the volcanic plume. Our analysis shows that values of the GPS residuals are ten times smaller than those found for the Miyakejima eruption (Japan), highlighting a likely relationship between residuals and eruption intensity. In the future, data derived from the GPS stations located on Etna’s flanks could be used to improve the alerting system of volcanic ash, already operating at the Istituto Nazionale di Geofisica e Vulcanologia, Osservatorio Etneo.