The analysis of seismic data represents today the primary tool in the search for the presence of postperovskite in the lowermost mantle (D″). This work aims at testing whether the inversion of gravitational data can also contribute to the detection of postperovskite in D″. We assume that the transition from perovskite to postperovskite is accompanied by a reduction in viscosity and test the effects of such viscosity change on the prediction of the dynamic geoid with a numerical model of subducted lithosphere. Our results show that the long-wavelength component of the geoid is very sensitive to the presence of postperovskite areas in D″, especially if their viscosity is significantly lower than the viscosity of the surrounding perovskite and if these areas are located close to density anomalies, i.e. subducted slabs.