Grounding with Google Search

1 min read

Step 1: Enable grounding

Turn on Google Search or Google Maps grounding in the model configuration

Step 2: Model queries Google

The model issues one or more search queries to retrieve real time information

Step 3: Results incorporated

Search results are woven into the model's response with source citations

Step 4: Verify with sources

Clickable source links let users verify every grounded claim

Google's search advantage

Gemini is the only major AI API with direct access to Google's own search index. Competing APIs rely on Bing, Brave, or Tavily for web search. Google's index is the most comprehensive, meaning grounded responses from Gemini models tend to have broader and more up to date coverage.