On September 15, Google announced that Gemini 3.8 Live is rolling out in Search Live. It is the standard 3.8 Live model named for Search Live. Google’s Extended Thinking version has a separate rollout across the Gemini app, Workspace, developer tools and enterprise products.
Search Live is the voice-and-camera experience inside the Google app. Google expanded it in March to the languages and locations where AI Mode is available, then described access in more than 200 countries and territories. A person can ask a question aloud, follow up and open the web links shown with the answer.
The model changed. For a service business, the useful thing to test is the longer conversation Search Live already supports. A prospect can move from a broad problem to a specific decision without starting a fresh search each time.
The change is in the conversation, not a new ranking rule
Google’s announcement describes a new model and where it is rolling out. It does not announce a ranking update, a new citation rule or a shortcut for getting a business into Search Live.
Google says Gemini 3.8 Live can process visual input, switch among 97 supported languages during a conversation and keep talking while it runs tools in the background. Those are vendor claims about the model. The announcement does not say that every model ability appears the same way in Search Live or that any of them changes how a page ranks.
So start with observation. See what the product does with the questions that matter to your customers before you change a page or move a search budget.
Test the question chain your customers use
Pick one real task that often comes up before an inquiry. It could be a question about fit, timing, service area, price conditions or what somebody needs to prepare before booking.
Open Search Live in the Google app and run the task in three parts:
- Ask the first question the way a customer would say it aloud.
- Ask a follow-up that narrows the situation without repeating the whole query.
- Ask the decision question that would normally lead to a call, form or booking.
If the service involves something visible, run a second pass with the camera. Keep that pass separate so you can see what changed when visual context entered the conversation.
At the end, open the transcript. Record the question chain and every linked page. Also note any request for clarification. It can reveal a detail your own page needs to make clearer.
Record the handoff to the web
The spoken answer is only half the test. Look at what happens when Search Live sends the person to a website.
- Which pages did it link to during the first answer and the follow-ups?
- Did any linked page answer the decision question directly?
- Was the service area, timing, price condition or proof easy to find on a phone?
- Did the page offer a clear next action once the answer was useful?
One run is a sample, not a visibility report. Save the date, language, location and exact question chain. Repeat the same path later before you call a change a trend.
Fix the page gap before guessing at the model
If the test exposes a missing answer, repair that answer where a customer would expect to find it. Put timing beside the service it qualifies. Put service-area details near the booking step. Keep proof close to the claim it supports.
In Glory’s landing-page and CRM work, this is the useful handoff: shape the page around the real decision, make the next action clear and carry the inquiry into the backend system.
Then run the same spoken path again. You are looking for a better customer journey whether Search Live links to the page on that attempt or not.
Gemini 3.8 Live gives Search Live a new model for its conversation layer. The practical response is a small, dated test: one task, two follow-ups, the linked pages and the next action. That is enough to find the first gap worth fixing without turning a model release into a ranking theory.
Source note: Product and rollout details come from Google’s September 15 Gemini announcement and its March 26 Search Live expansion announcement. Model-performance statements are Google’s claims. Glory wrote the testing method and recommendations.