Gemini API gets live Google Search grounding

Ask an LLM who won a tournament this weekend, and you usually get an apologetic note about knowledge cutoffs or a completely confident hallucination.

Google is tackling that headache directly. The company just rolled out Grounding with Google Search across Google AI Studio and the Gemini API, letting developers tether Gemini models to real-time search data.

The feature supports all generally available versions of Gemini 1.5. Along with up-to-date answers, responses return in-line source links and search suggestions pointing users back to relevant search results.

Real-time web access at a price

Turning search on inside Google AI Studio is as simple as flipping a toggle under Tools, where testing is free. In the API, developers enable the google_search_retrieval tool.

Live web retrieval does not come cheap, however. In the paid API tier, Google charges $35 per 1,000 grounded queries, on top of normal token processing.

To keep costs and latency under control, Google introduced dynamic retrieval. The system scores incoming prompts from 0 to 1 based on how likely they are to need web search.

Developers can configure a custom threshold, which defaults to 0.3. If a user simply asks for code formatting or creative brainstorming, Gemini skips the live search call and saves your budget.

Later updates expanded availability across Europe, letting developers there test it for free in AI Studio and run production workloads through the paid tier.

Why it matters

For developers building research assistants, market trackers, or customer-facing bots, this removes the need to build and maintain bespoke web scraping and RAG pipelines just to get today's news. Getting structured citations and search suggestions right out of the box makes building trustworthy chatbots much simpler.

At $35 per 1,000 queries, however, calling live search indiscriminately will drain an API budget very quickly. We think dynamic retrieval is the smartest part of this rollout because it treats web access as an expensive tool rather than an automatic default.

Search grounding turns Google's biggest structural moat into an API primitive. While competitors have to rely on third-party search indexes or complex multi-agent setups, Google can plug its own world-class index straight into Gemini, setting a tough benchmark for factual accuracy in agentic workflows.