The transition towards autonomous public transport (APT) is considered a key lever for achieving ecological and social sustainability. However, passenger acceptance is a critical prerequisite for the successful deployment of APT, particularly given the absence of a human driver. A potential to foster acceptance is the use of virtual in-vehicle assistants (VIVAs), as they offer a promising means of compensating for a possible lack of communication, support, and passengers’ subjective sense of safety, particularly in situations involving unexpected system behaviour. This study addresses this potential by developing and evaluating concrete design features (DFs) that operationalize existing design principles (DPs) for comprehensive VIVAs in APT. Following a design science research approach, the DFs were validated through an iterative evaluation process including six think-aloud sessions with experts and potential users. The study identifies in sum 32 DFs, 17 fundamental DFs and 14 supplemental DFs, emphasising personalized yet restrained communication, transparency to foster trust, clear emergency guidance, adaptive support for vulnerable users, and cybersecurity measures. A key finding is the distinction between fundamental and supplementary DFs, with users consistently favoring functionally simple and non-intrusive designs. This work contributes prescriptive and reusable design knowledge by systematically translating abstract DPs into actionable DFs, thereby bridging the gap between theory and implementation.