<p>Virtual assistants have their origins in the first efforts to emulate human intelligence through systems capable of passing the Turing test. Over the years, they have evolved significantly, with Siri, introduced in 2011, being the first system explicitly recognized as a virtual assistant. Since then, the development of virtual assistants has spread to various disciplines, with the aim of assisting one or more users in various tasks, using increasingly sophisticated technologies. This article presents a systematic review, in which 34 scientific literature reviews, and surveys published between 2015 and 2025 were identified and analyzed, in order to delimit the concept of virtual assistant, its evolution and current trends, including the impact of technologies such as large-scale language models. The findings reveal a terminological ambiguity, a transition from rule-based and closed-domain implementations to Artificial Intelligence driven assistants, and the emergence of capabilities such as personalization, proactivity, context awareness, and multimodal interaction. The review also synthesizes recurring limitations and challenges reported in the last decade, highlighting patterns across domains and temporal shifts. These insights contribute to a clearer conceptualization and outline promising directions for future research.</p>