The Self-Prioritisation Effect (SPE) reflects the ability to efficiently discern self-relevant information. The self-voice emerges as a crucial identity marker, due to its inherent self-relevance, and previous work has demonstrated the perceptual and cognitive advantages of the self-voice over other voices. Yet, the extent to which humans prioritise their self-voice when they hear it is because it is both self-similar (“That sounds like my voice”) and self-generated (“I said that”) remains understudied. Here, we examined impacts of self-similarity and self-generation on the SPE through 3 experiments. In each experiment, participants learned associations between three voices and three identities (self, friend, other), then performed a task requiring them to perceptually match the heard voices with visual labels (“you”, “friend”, “stranger”). Experiment 1 revealed an augmented SPE when the self-associated voice in the task was the participant’s own self-similar and self-generated voice. In Experiment 2, the SPE was diminished when the self-voice was associated with the “stranger” label – here, the other-associated, but self-similar and self-generated, voice was similarly prioritised to a self-associated but unfamiliar voice. In Experiment 3 we investigated the role of self-generation, by associating the self with a self-similar but machine-generated audio clone of the participant. The SPE was again enhanced. In sum, we demonstrate that listeners show flexibility in their mental representation of self, where multiple sources of self-related information in the voice can be jointly and severally prioritised, independently of self-generation. These findings have implications for the application of self-voice cloning within voice-mediated technologies.