When a voice assistant may be a partner or a tool: Older adults’ context-dependent preference patterns — A systematic review

Background Intelligent voice assistants (IVAs) hold particular promise for supporting older adults in daily activities, yet the research landscape across diverse application scenarios remains fragmented. Objective This systematic review synthesizes 113 peer-reviewed publications to examine IVA use in older adults’ daily lives, covering support strategies, acceptance factors, user preferences, and emerging research directions. Methods Following PRISMA 2020 guidelines, five core databases were systematically searched, supplemented by Google Scholar and citation chasing. Two independent reviewers conducted screening, data extraction, and thematic coding ( κ = 0.86). Results The literature reveals a clear thematic evolution from basic acceptance and feasibility studies toward emotional companionship and social connection research. Six support domains were identified: life assistance, healthcare management, social independence, emotional support, learning, and travel assistance. Acceptance factors were organized into user-related, IVA-related, and social categories. An emerging pattern across heterogeneous studies suggests context-dependent preferences among older adults: they tend to favor partner-like, emotionally warm interaction styles in social and companionship contexts, while preferring more professional, tool-oriented roles for functional and health-related tasks. This pattern remains to be confirmed by direct within-participant comparisons across contexts. Critical evidence gaps include limited research on complex multi-step activities, a scarcity of longitudinal studies with objective outcomes, and underexplored cross-device role consistency. Conclusion IVA research is shifting toward emotional and social engagement paradigms. Addressing the identified gaps through longitudinal designs and adaptive persona architectures is essential for developing genuinely age-friendly voice companions.