Perplexity Launches Voice Assistant Mode

Voice Mode Launch

Perplexity launched Voice Assistant Mode in March 2026 for iOS and Android, available to all users including the free tier with usage limits, and unlimited for Pro subscribers. The feature used a combination of custom speech-to-text (optimised for query-style utterances rather than dictation), Perplexity's search and synthesis engine, and text-to-speech for response delivery — with the full interaction completing in typically 2-4 seconds including web search.

Low Latency Design

Perplexity's engineering team optimised specifically for response latency — the time between the user finishing speaking and the AI beginning to respond. By parallelising speech recognition, query classification, and initial search initiation, they achieved median response latency of 480ms — competitive with purpose-built voice assistants and significantly faster than general-purpose AI models adding voice interfaces.

Conversational Follow-ups

Voice mode maintained conversational context across a session, allowing natural follow-up questions without restating the subject. "What's the capital of India?" followed by "What's its population?" correctly resolved "its" to Delhi. This context persistence made voice interactions feel genuinely conversational rather than a sequence of isolated queries — a significant UX improvement over assistant voice modes that reset context after each exchange.

Web Grounding for Voice

The key differentiator was that all Perplexity voice responses were grounded in real-time web data — not a language model's parametric knowledge. Asking "What's happening in the stock market today?" returned information from live financial sources rather than the model's training data. For time-sensitive information queries, this real-time grounding was a fundamental advantage over all major voice assistants.

What This Means for Indian Businesses