Probabilistic seismic hazard assessment relies on long‐term earthquake forecasts and ground‐motion models. Our aim is to improve earthquake forecasts by including information derived from geodetic measurements, with an application to the Colombia–Ecuador megathrust. The annual rate of moment deficit accumulation at the interface is quantified from geodetically based interseismic coupling models. We look for Gutenberg–Richter recurrence models that match both past seismicity rates and the geodetic moment deficit rate, by adjusting the maximum magnitude. We explore the uncertainties on the seismic rates (‐ and ‐values, shape close to ) and on the geodetic moment deficit rate to be released seismically. A distribution for the maximum magnitude bounding a series of earthquake recurrence models is obtained for the Colombia–Ecuador megathrust. Models associated with values compatible with the extension of the interface segment are selected. We show that the uncertainties mostly influencing the moment‐balanced recurrence model are the fraction of geodetic moment released through aseismic processes and the form of the Gutenberg–Richter model close to . We combine the computed moment‐balanced recurrence models with a ground‐motion model, to obtain a series of uniform hazard spectra representative of uncertainties at one site on the coast. Considering the recent availability of a massive quantity of geodetic data, our approach could be used in other well‐instrumented regions of the world.