Displacement experiments have demonstrated that experienced migratory birds translocated thousands of kilometers away from their migratory corridor can orient towards and ultimately reach their intended destinations [1]. This implies that they are capable of “true navigation”, commonly defined [2–4] as the ability to return to a known destination after displacement to an unknown location without relying on familiar surroundings, cues that emanate from the destination, or information collected during the outward journey [5–13]. In birds, true navigation appears to require previous migratory experience [5–7, 14, 15, but see 16, 17]. It is generally assumed that, to correct for displacements outside the familiar area, birds initially gather information within their year-round distribution range, learn predictable spatial gradients of environmental cues within it and extrapolate from those to unfamiliar magnitudes ̶ the gradient hypothesis [6, 9, 18–22]. However, the nature of the cues and evidence for actual extrapolation remains elusive. Geomagnetic cues (inclination, declination and total intensity) provide predictable spatial gradients across large parts of the globe and could serve for navigation. We tested the orientation of long-distance migrants, Eurasian reed warblers, exposing them to geomagnetic cues of unfamiliar magnitude encountered beyond their natural distribution range. The birds demonstrated re-orientation towards their migratory corridor as if they were translocated to the corresponding location but only when all naturally occurring magnetic cues were presented, not when declination was changed alone. This result represents direct evidence for migratory birds’ ability to navigate using geomagnetic cues extrapolated beyond their previous experience.