Having recognized gender bias as a major issue affecting current translation\ntechnologies, researchers have primarily attempted to mitigate it by working on\nthe data front. However, whether algorithmic aspects concur to exacerbate\nunwanted outputs remains so far under-investigated. In this work, we bring the\nanalysis on gender bias in automatic translation onto a seemingly neutral yet\ncritical component: word segmentation. Can segmenting methods influence the\nability to translate gender? Do certain segmentation approaches penalize the\nrepresentation of feminine linguistic markings? We address these questions by\ncomparing 5 existing segmentation strategies on the target side of speech\ntranslation systems. Our results on two language pairs (English-Italian/French)\nshow that state-of-the-art sub-word splitting (BPE) comes at the cost of higher\ngender bias. In light of this finding, we propose a combined approach that\npreserves BPE overall translation quality, while leveraging the higher ability\nof character-based segmentation to properly translate gender.\n