By using Google’s search results to ground generative AI applications, developers can provide users with more accurate, relevant, and trustworthy information, the company said.
Google Cloud has made Grounding with Google Search available in the Google AI Studio and in the Gemini API. The new feature enables developers to get more accurate and fresh responses from Gemini generative AI models, the company said.
In addition to more accurate responses, the model returns links to the grounding sources and search suggestions that point users to search results that correspond to the grounded responses.
Grounding with Google Search is supported for all generally available versions of Gemini 1.5 models, Google Cloud said. It can be accessed in Google AI Studio under the “Tools” section or in the Gemini API by enabling the google_search_retrieval tool. Users can test grounding for free in Google AI Studio. With the API, developers can access the tool with the paid tier of $35 per 1,000 grounded queries.
Grounding benefits include the following:
- Grounding helps to ensure that AI applications provide users with more factual information.
- Models using grounding can access real-time information, thus making AI applications relevant and applicable to a wider range of scenarios.
- By providing supporting links, grounding brings transparency to AI applications.
- By drawing information from Google Search to enhance the model response, grounding is able to offer richer color on many queries.
The way that Grounding with Google Search works is when a user makes a query with grounding turned on, the service uses Google’s search engine to find up-to-date and comprehensive information relevant to the query, and sends it to the model. The model then responds with higher accuracy and freshness, providing inline grounding sources and search suggestions.
