Gaze-cueing is the automatic spatial orienting of attention in the direction of perceived gaze. Participants respond faster to targets located at positions congruent with the direction of gaze, compared to incongruent ones (gaze cueing effect, GCE). However, it still remains unclear whether its occurrence depends on intact integration of information from the entire eye region or face, rather than simply the presence of the eyes per se. To address this question, we investigated the GCE in PS, an extensively studied case of pure acquired prosopagnosia. In our gaze-cueing paradigm, we manipulated the duration at which cues were presented (70ms vs. 400ms) and the availability of facial information (full-face vs. eyes-only). For 70ms cue duration, we found a context-dependent dissociation between PS and controls: PS showed a GCE for eyes-only stimuli, whereas controls showed a GCE only for full-face stimuli. For 400ms cue duration, PS showed gaze-cueing independently of stimulus context, whereas in healthy controls a GCE again emerged only for full-face stimuli. Our findings suggest that attentional deployment based on the gaze direction of briefly presented faces requires intact processing of facial information, which affords salience to the eye region.