AI search · Glossary
What is AI grounding?
Grounding is when an AI assistant bases its answer on sources it retrieves at the time, such as web search results, instead of only on what it learned in training. Grounded answers usually show citations to the sources used.
Why Grounding matters
Grounding is how current information about your business reaches AI answers. Microsoft says Copilot is grounded in web search results and shows linked citations, and Bing Webmaster Tools reports the grounding queries that led its AI features to your pages.
How Grounding works
- The assistant turns the user's question into one or more search queries.
- It reads the results and writes an answer based on them.
- It links the pages it relied on, which is where citations come from.
How to apply it
- Check the grounding queries in Bing Webmaster Tools' AI Performance report.
- Make sure assistants' search crawlers can reach your pages.
- Write sections that answer one question each, so a retrieved passage makes sense on its own.
Example
Bing Webmaster Tools shows a dental practice that a grounding query about Sunday emergency appointments cites its hours page, so the practice keeps that page's hours current.
Common mistake
Assuming grounding makes answers accurate. OpenAI says search results and citations can be incomplete, outdated or incorrect.
put this into practice