Detection of gravity change over time has been used to better understand magmatic activity at volcanoes for decades, but the technique is not commonly applied to forecasting eruptions. In contrast, other tools, notably seismic, deformation, and gas monitoring have made exceptional strides in the past several decades and form the foundation for eruption forecasting, especially during the final buildup to an eruption. Reasons for this gap include the high cost and fragile nature of gravity instruments, and the ambiguous nature of many results. But this is changing. Instrumentation is becoming more robust and accurate, expenses may soon diminish thanks to technological advances, and the record of success in tracking subsurface mass change (either from magma or hydrothermal fluids) in volcanic areas is growing. Here we review how gravity change can be applied to forecasting volcanic eruptions across a variety of spatial and temporal scales. We argue that microgravity has untapped potential as a forecasting tool in three specific ways: constraining probabilistic assessments, detecting long-term mass change that may occur prior to the onset of vigorous seismicity and deformation, and identifying transient activity that indicates magma ascent or other changes that immediately precede new eruptions or changes in ongoing eruptions. As with any volcano-monitoring method, microgravity has strengths and weaknesses, but the varied forms of data collection—for instance, campaign versus continuous, and relative versus absolute—offer the potential to record a broad range of signals at volcanoes with a diversity of magmatic systems. The technique is currently underutilized; additional attention, investment, and application at more volcanoes could help to realize its promise.