Site-specific force measurements on a rutile TiO 2 (110) surface are combined with first-principles calculations in order to clarify the origin of the force contrast and to characterize the tip structures responsible for the two most common imaging modes. Our force data, collected over a broad range of distances, are only consistent with a tip apex contaminated with clusters of surface material. A flexible model tip terminated with an oxygen explains the protrusion mode. For the hole mode we rule out previously proposed Ti-terminated tips, pointing instead to a chemically inert, OH-terminated apex. These two tips, just differing in the terminal H, provide a natural explanation for the frequent contrast changes found in the experiments. As tip-sample contact is difficult to avoid while imaging oxide surfaces, we expect our tip models to be relevant to interpret scanning probe studies of defects and adsorbates on TiO 2 and other technologically relevant metal oxides.