Predicting earthquakes is a long-desired goal. The main challenge is to identify precursory signals that reliably predict the impending earthquake. Since hydrological and hydrogeochemical properties and processes can be very sensitive to minute strains, the hope is that measurements from hydrological systems might record precursory rock deformation that would otherwise be undetectable. Of the many hundreds of studies, we review a subset to illustrate how signals can be challenging to interpret and highlight questions raised by observations—examples come from China, Japan, Taiwan, India, the USA, Russia, France, Italy and Iceland. All are retrospective studies. Some signals seem to have no other explanation than being precursory, however, rarely is enough data available to undertake a thorough analysis. Some hydrological precursors might be recording deformation events that are slower than traditional earthquakes (and hence usually harder to detect). Long times series of data are critical for both identifying putative precursors and assessing their origin and reliability.