Grounding

Grounding is how an answer engine ties what it writes to evidence. Google's Gemini docs describe a Grounding with Google Search feature that lets the model draw on search results, and the response carries metadata linking each part of the answer to its source: groundingSupports "provides specific mappings that link the generated content to its relevant citation source in the chunks."

Grounding is not automatic. Anthropic's docs say "Claude determines when to search based on the prompt", and that it searches when a request depends on information that is "current, changing, or outside its training data". OpenAI's docs say the model "can choose to search the web or not based on the content of the input prompt."

For you, that means grounded answers are where citations happen. Questions about prices, policies, recent releases and specific products are the ones most likely to trigger a search, so those are the pages most worth keeping accurate, crawlable and easy to quote. A broad, stable question may be answered from training data alone, with no links for anyone.

Related terms

• Retrieval-augmented generation
• Citation
• Query fan-out
• Answer engine

Sources

1. Grounding with Google Search (opens in a new tab) · Google Cloud · Updated 28 Oct 2026
2. Web search tool (opens in a new tab) · Anthropic · Checked 28 Oct 2026
3. Web search (API guide) (opens in a new tab) · OpenAI · Checked 28 Oct 2026