Grounding Query

What is a grounding query?

A grounding query is the search query an AI system runs internally to retrieve supporting material before generating an answer. It is not the question the user typed into the chat box, but the reformulated query the AI sent to a search engine in order to answer it. If grounding is the process of anchoring an answer to external evidence, the grounding query is the first step of that process — what the evidence was searched with.

It differs from the user's question

When a user asks "is there a trustworthy tax accountant near me?", the AI does not search that sentence verbatim. It reformulates the request into several search queries combining location, service type, and evaluation criteria, then selects source documents from the results. The "user keywords" shown by search term tools and the "grounding queries" shown by AI citation measurement tools are therefore different data — and the actual path through which your content gets surfaced is closer to the latter.

Where you can see them

As of 2026, two notable free tools expose grounding queries.

Both are Microsoft-side aggregations, and neither publishes a list of which AI platforms are covered — a limitation to keep in mind.

Measurement takeaways

• A direct clue for content alignment — knowing which grounding queries surface your pages tells you which phrasings your headings and direct-answer paragraphs should match. This is closer to the AI retrieval path than optimizing for user keywords.
• Citation and mention are separate — being retrieved and cited via a grounding query does not mean your brand is named in the answer body. What happens after the citation is covered under ghost citation.
• Read it together with share — how much of a given grounding query's citations your domain captures is what Citation Share measures.

Related terms

• Grounding
• Citation Share
• AI Retrievability
• Ghost Citation
• Generative Engine Optimization (GEO)