The removal of free-surface-related multiples plays an important role in seismic data processing. We have developed a novel scheme for predicting free-surface-related multiples by combining the revised Marchenko equations with free-surface effects and convolutional seismic interferometry. This data-driven method can create the free-surface-related multiple prediction using only the reflection response recorded at the free surface and a macrovelocity model estimated in the first layer. By setting only one reference boundary in the first layer, we can stably and efficiently retrieve the one-way upgoing and scattered downgoing Green’s functions at virtual focusing points and then predict all orders of free-surface-related multiples. Numerical experiments on a 2D model and a subset of the Marmousi model indicate that this method can accurately and efficiently predict the traveltimes of all orders of free-surface-related multiples. After applying least-squares matched filtering, all orders of the surface-related multiples can be effectively eliminated from the raw shot gathers.