Volcanoes often show subtle warning signs before they erupt, but detecting these signals in time is a major scientific challenge. In this study, we introduce a new approach of geometric phase to track underground changes at volcanoes, which we extract from background ground motion signals in the Earth known as ambient seismic noise. Unlike traditional techniques that focus on changes in wave travel times, our method looks at the wavefield geometriy captured by a network of sensors. We applied this technique to Hawai‘i’s Kīlauea volcano and found it could clearly detect key events before, during, and after the major 2018 eruption—including magma movements and the collapse of the volcanic crater. It also picked up signs of later volcanic activity between 2020 and 2024. Because of the computaitonal speed, reliability, and ability to detect early warning signs of the geometric phase sensing, we believe this geometric phase technique could be useful far beyond volcanoes—for example, in monitoring glaciers, groundwater, ecosystems, and oceans. Future research can build on this by combining it with other environmental data and applying it in more places around the world.