Google brings Gemini 3.8 Live to Search Live for near real-time voice search

What shipped

Google has added a new conversational model, Gemini 3.8 Live, to Search Live, the voice-driven side of its search product. According to a dev.to report summarizing Google's announcement, the model handles spoken dialogue with reasoning that happens close to real time, and it can take visual context — what the person is looking at on screen — into account while the exchange is still running.

The result is a search experience that behaves more like an ongoing conversation than a sequence of typed queries and static results pages. As the dev.to report describes it, Search Live supports interruptions, language switching partway through a session, and follow-up questions that narrow or redirect a task without the user having to restart from scratch.

Two variants for two jobs

Google is shipping the model in two flavors:

• Gemini 3.8 Live, tuned for live conversation at scale with an emphasis on cost efficiency. Google says it can run tools and trigger API calls in the background while the dialogue continues.
• Gemini 3.8 Live Extended Thinking, aimed at harder problems that need deeper, multi-step reasoning.

Both versions are available in Search Live, according to the report, and both are being pushed out across other parts of Google's ecosystem.

Where it is available

Beyond Search Live, Google says the models are reaching the Gemini app, Google Workspace, developer tools and enterprise products. Developers can experiment with Gemini Live through the Gemini Live API and Google AI Studio, while some enterprise channels are being handled through private previews and integrations Google says are still coming.

The announcement also covers audio provenance: Google says it applies SynthID watermarking to audio output as a transparency measure. That detail grows in importance as spoken interfaces spread, since generated speech is far harder to verify by ear than text on a page.

One thing the announcement does not cover is cost. According to the dev.to report, no pricing or API pricing changes were stated, so anyone planning to build on the model is waiting on commercial terms.

What it changes for search

Conspicuously absent from the material is any new ranking framework or a dedicated Search Live optimization program. Google has not, on the strength of this rollout, rewritten the SEO rulebook. What changes instead is the shape of the interaction: people speaking naturally, adding a detail after glancing at something on screen, drifting between languages, and chaining follow-ups in ways that don't map onto a tidy keyword funnel.

The dev.to report singles out multilingual switching as especially relevant for organizations that serve audiences in multiple languages. If a user can slip from one language to another inside a single conversation, translated pages stop being an afterthought — factual consistency and well-maintained localized information become part of whether a site holds up under conversational questioning.

For developers, the more interesting signal may be the plumbing. A model that keeps a conversation going while it quietly executes tools and API calls in the background is the raw material for genuinely agentic voice experiences, and Google is exposing it through its API and AI Studio rather than keeping it locked inside Search.

Why it matters

This is Google repositioning search itself around speech. If follow-up questions, interruptions and language switches become normal inside Search Live, the unit of search shifts from one query to one session — and content that answers plainly, keeps its facts current and survives being quoted aloud in fragments will fare better than pages built purely around keyword matching.

It also matters competitively: near real-time voice search with live reasoning is the capability that separates an assistant from a search box, and Google is now shipping it into its highest-traffic product. The details here come from a single dev.to report on Google's announcement, so specifics such as availability windows and pricing should be treated as subject to confirmation — but the direction is clear: conversation is becoming the default interface for finding information.