The western Pacific hosts major subduction systems such as Izu–Bonin–Mariana and Tonga–Kermadec, but also less conspicuous systems such as Yap, Mussau and Hjort trenches which constitute the young, incomplete, or ultraslow-member in the evolutionary spectrum of subduction zones. We used satellite-derived gravity data to compare well-developed and immature subduction systems. It is shown that at spatial resolution > 10–20 km or so, the satellite data have accuracy comparable to ship-board gravity measurements over intra-oceanic subduction zones. In the isostatic residual gravity anomaly map, the width of non-isostatically-compensated region of the mature subduction zones is much wider than that of immature ones. More importantly, when the gravitational attraction due to seafloor is removed, a large difference exists between the mature and immature subduction zones in the overriding plate side. Mature subduction zones exhibit broad low gravity anomalies of ~ 200–250 mGal centered at distances of 150–200 km from the trench which are not found over immature subduction zones. The cause of the broad low gravity anomalies over mature subduction zones is debatable due to lack of information on the deep crust and upper mantle structure and property. We discuss the following four causes: (1) serpentinization of the upper mantle beneath the forearc; (2) presence of partial melt in the mantle wedge caused by release of volatiles from the slab, frictional heating and distributed by mantle circulation; (3) difference in density structure between the overriding and subducting plates caused by difference in age and thermal structures with and without compositional stratification between crust and mantle; and (4) anomalous thickness of the arc not explained by isostasy. Our analysis suggests that serpentinization cannot explain the observed gravity anomaly which appears ~ 150–200 km from the trench. Although the extent and distribution of partial melt within the mantle wedge remain in question, to our best estimate, partial melting contributes little (< 50 mGal) to the total negative gravity anomaly. The difference in density structure reflecting temperature difference can only explain less than half of the low gravity anomaly. The sinking of lighter crustal material produces a large negative anomaly in the forearc but its location does not match the observed gravity anomaly. It appears that one cannot explain the total difference in gravity anomaly without invoking anomalous thickness of the arc. Although we could not identify the sole or combination of factors that give rise to the low gravity anomaly in mature subduction zones, the comparison of gravity anomalies between mature and immature subduction zones is likely to provide an important constraint for understanding the evolution and structure of subduction zones as more complementary evidences become available.