In Italy, volcanic earthquakes are crucial for assessing seismic hazard, as several volcanoes pose significant risks to densely populated areas (e.g., Catania and its surroundings, Phlegraean Fields, Vesuvius). Recently, the increase in seismic activity in the Phlegraean Fields, characterized by swarms of low-magnitude earthquakes concomitant with bradyseism, has prompted a revision of the configuration of the Ground Motion Models (GMMs) used in the U.S. Geological Survey (USGS) ShakeMap software implemented in Italy for volcanic areas. In this study, we applied a cross-validation technique to evaluate the effectiveness of various GMMs available in the literature for volcanic regions, aiming to the identification of the most appropriate configuration to update and improve the ShakeMap service provided by the Istituto Nazionale di Geofisica e Vulcanologia (INGV). We compared the performance of the currently adopted GMMs [a combination of Tusa and Langer (2016) for earthquakes down to 5 km depth and Bindi et al. (2010) for deeper earthquakes] in predicting ground shaking with those developed by Tusa et al. (2020) and Lanzano and Luzi (2020). To conduct the tests, we used 100 Italian earthquakes that occurred in volcanic areas between February 2019 and May 2024, with magnitudes in the range 3.0‑4.5 and depths ≤ 35 km. Among the tested models, the GMM proposed by Tusa et al. (2020) was found to provide the most accurate predictions and will be adopted for generating the INGV shakemaps.