Women who have experienced gender-based violence (GBV) are at an increased risk of developing mental illnesses such as depression, anxiety, and post-traumatic stress disorder (PTSD). Recently, Artificial Intelligence (AI) has provided new tools to assist mental health clinical diagnosis, including speech-based detection. However, there is not much work done on the GBV victim (GBVV) condition detection. This study aims to identify specific speech features that aid this detection, analyse the relationship of such results with the user's psychological evaluation, and evaluate whether the models rely on the speaker identity or self-reported emotions to predict the GBVV condition. Our results indicate that it is possible to distinguish GBVV with controlled sequelae from non-victims, which may suggest that such differentiation for GBVV with more severe mental aftereffects-such as PTSD-may be even more meaningful. We believe that our work can help future mental health AI therapy assistants.