Large-scale land change maps are essential to support policies addressing land transformations. Calibration and validation of large-scale land change maps require reference data that are commonly acquired by visual interpretation of remotely sensed images. However, visual interpretation itself is prone to error. Little is known about factors influencing the quality and consistency of changes detected by visual interpretation. This paper reports on an experiment assessing the effect of the number of very high resolution images and land change process types on the consistency of visual interpretations. The experiment involved 48 sites scattered over Europe for which 18 individuals interpreted very high resolution images, which were provided via Google Earth. Land change process type was found to have a significant impact on the consistency of visual interpretations, while the marginal effect of the number of images was not significant. Absence of change on non-agricultural land was interpreted with high consistency. On the contrary, agricultural land abandonment and reforestation were the least consistent in their interpretation. We conclude that for increased efficiency, resources allocated to acquire reference data by visual interpretation should be adjusted based on the expected type of land change. Interpretation of agricultural land abandonment, reforestation and agricultural land expansion require most efforts.