Google has begun rolling out “Search Live in AI Mode” as a limited, opt-in experiment in the United States for both Android and iOS users. First demonstrated at the I/O 2025 conference, the feature enables real-time, back-and-forth voice conversations directly within the Google app, powered by a custom-engineered version of the Gemini AI model. Users can access the function by tapping a new waveform icon beneath the search bar, which opens a full-screen conversational interface.
This interface allows for natural language follow-up questions without restarting the search. It includes a mute function and a transcript option, which provides a real-time text version of the dialogue and allows for typed interaction. As the conversation progresses, the AI displays relevant website links on screen. To facilitate multitasking, the feature can continue operating in the background even if the user minimizes the app. All dialogues are automatically saved to the AI Mode history for future reference.
Technologically, Search Live utilizes a “query fan-out” method, where the AI expands its search to gather information from a wide array of sources for a more comprehensive response. Future updates are planned to integrate camera-based object recognition, allowing users to point their device at an object and ask questions about it verbally. For personalization, users can currently choose from four distinct AI voices: Cosmo, Neso, Terra, and Cassini.
