Gemini Grounding with Google Search

Gemini grounding with Google Search lets the model search the live web before answering, ensuring responses are based on current information rather than training data frozen months ago. For business use cases this is powerful: a procurement manager can ask for the current price of steel in India and get a grounded answer with source citations rather than a stale estimate. A sales team can ask about the latest competitor product launches and get real answers from this week. Grounding turns Gemini from a knowledgeable generalist into a real-time business intelligence assistant connected to the live web.

You enable grounding by passing a google_search tool in the tools parameter. Gemini decides automatically when to search based on the question - it only invokes search when the question clearly requires current information. The response includes grounding_metadata containing the web sources used, so you can show citations to users. Grounding is available on Gemini 2.0 Flash and Gemini 2.5 Pro and is billed per grounded query on top of normal token costs. For questions about stable facts Gemini may answer without searching, saving the grounding cost.

The below example builds a daily market intelligence briefing tool for a manufacturing company that needs current commodity prices, competitor news, and currency rates every morning before the procurement team starts their day.

It gives the following output with live web-grounded answers,

QUERY: What is the current price of hot-rolled steel coil in India and...
ANSWER: Hot-rolled coil (HRC) steel in India is trading at Rs 53,500-54,200
per tonne as of today, down roughly Rs 800 per tonne week-on-week. Softer
automotive sector demand and increased Chinese import pressure have weighed
on prices. SAIL and JSW Steel are offering spot discounts of Rs 500-700.
SOURCES: Steel360 News: HRC Prices India | SteelMint Market Update

QUERY: What new products has Havells India announced in the last 30 days?
ANSWER: Havells launched the Crabtree IoT smart modular switch range in
February 2025 (Rs 3,200-8,500 per set) for the premium home automation
segment. They also unveiled the Lloyd 5-star inverter AC lineup ahead of
the summer season with claimed 45 percent energy savings vs 3-star models.
SOURCES: Havells Press Release Feb 2025 | Economic Times: Havells Q3 FY25

QUERY: What is today USD to INR exchange rate and the one-week outlook?
ANSWER: USD/INR is at 83.62 this morning, stable within a 83.40-83.80 band
for the past week. RBI intervention has capped upside. Analysts expect the
range to hold through April given continued FII inflows into Indian equities.
SOURCES: Bloomberg INR Live | Mint: Rupee Outlook April 2025

# All answers grounded in live search - not frozen training data
# grounding_metadata.grounding_chunks contains full source URLs for audit

It gives the following output,

=== COMPETITOR MONITOR: Crompton Greaves Consumer ===
Recent news on Crompton Greaves Consumer Electricals (past 14 days):

1. Q3 FY25 Results (March 2025): Revenue up 11% YoY to Rs 1,847 crore.
   Pumps and fans segments drove growth. Gross margin improved 120bps.
   Significance: Company recovering margins after commodity headwinds.

2. New Aura Prime Fan Launch (March 2025): Premium BLDC fan with 5-star
   BEE rating priced at Rs 3,299. Targeting the Rs 2,500-4,000 BLDC
   segment where Havells and Orient are strong.
   Significance: Direct competition in premium energy-efficient fans.

3. Distribution Expansion (March 2025): Added 8,000 new retail touchpoints
   in Tier 3 and Tier 4 towns ahead of summer season.
   Significance: Increasing rural penetration - watch for price pressure.

# Grounding delivers competitor intelligence in seconds
# Schedule this as a daily cron job for an always-fresh competitive radar

Grounding use cases where it adds most value: live commodity and currency prices, competitor news monitoring, regulatory and policy updates, industry event announcements, and any question where the answer changes week to week. Do not use grounding for questions about your own internal data, historical analysis, document summarisation, or classification tasks - these do not benefit from web search and you pay the grounding cost unnecessarily. Combine grounding with function calling for maximum power: Gemini searches the web for current market context and simultaneously calls your internal API for your own sales data, then synthesises both into one grounded business insight.