Can social gaze behaviour reveal the leader during real-world group interactions? To answer this question, we developed a novel tripartite approach combining i) computer vision methods for remote gaze estimation, ii) a detailed taxonomy to encode the implicit semantics of multi-party gaze features, and iii) machine learning methods to establish dependencies between leadership and visual behaviours. We found that social gaze behaviour distinctively identified group leaders. Crucially, the relationship between leadership and gaze behaviour generalized across democratic and autocratic leadership styles under conditions of low and high time-pressure, suggesting that gaze can serve as a general marker of leadership. These findings provide the first direct evidence that group visual patterns can reveal leadership across different social behaviours and validate a new promising method for monitoring natural group interactions.