We performed a direct comparison of deuterium retention and release from tungsten in presence or in absence of oxygen impurities. A single crystal of W(110) was used to prepare tungsten with four different surface states: with its native oxide, atomically clean, covered with half a monolayer of oxygen atoms, and covered with three fourths of a monolayer of oxygen atoms. For a D ion fluence of 3×1021 D+ m-2 implanted at 300 K, deuterium retention was highest with the native oxide, lowest with three fourths of a monolayer of oxygen atoms at the surface and intermediate for the clean surface. This counterintuitive result is explained by a different localization of deuterium retention in these samples. For tungsten with its native oxide, deuterium retention occurs solely in the bulk, i.e. below the first atomic plane of the surface. For clean tungsten, deuterium retention occurs in part at the surface and sputtering should play a role. For tungsten with a sub-monolayer surface coverage of oxygen atoms, a transition from surface to bulk retention is observed above half a monolayer of adsorbed oxygen. Striking differences in desorption peak(s) temperature(s) are observed between D ion-implanted samples and D2 molecules-exposed samples. These results highlight the importance of the (near-) surface localization of oxygen and deuterium on the temperature dependence of deuterium desorption rate i.e. on the fusion fuel recycling coefficient.