Why does Copilot cite my page for queries I never targeted?

The query that retrieves your page is usually not the query a person typed. AI search systems rewrite a user's question into their own search phrases, and Microsoft calls the ones that led to a citation grounding queries. It describes them as the key phrases the AI used when retrieving content that was referenced in AI-generated answers.

Other engines document the same pattern. OpenAI says ChatGPT search typically rewrites a query into one or more targeted queries before sending them to search providers. Google says AI Mode may use query fan-out, issuing several related searches across subtopics. So the AI is not matching your keyword list. It is matching topics it decided were relevant to the question.

That explains the surprise. Your page gets pulled in whenever it fits a subtopic the system generated, even if you never wrote for that phrase.

Say you run a payroll software company and target "payroll software for small business". Copilot might cite your pricing page for a grounding query like "payroll processing cost per employee". You never targeted it. Your page simply holds the answer to a subtopic the AI searched for.

If you want the longer version of why the typed prompt and the retrieval query differ, see whether AI search uses the exact question a user types.

Grounding queries live in the AI Performance report in Bing Webmaster Tools. Microsoft opened it as a public preview in February 2026 and expanded it in June 2026. It covers Microsoft Copilot, AI-generated summaries in Bing, and select partner integrations.

The report shows:

• Total Citations and Average Cited Pages for your site
• Page-level citation activity, so you can see which URLs get cited
• Grounding queries, the phrases the AI used to retrieve your cited content
• Intents and Topics, which group that activity
• Citation Share and a Compare view

As of the August 2026 evidence review, none of this was available through the Bing Webmaster API. You read it in the interface or export it by hand. Our guide to exporting and analyzing Bing grounding queries walks through that step by step.

This matters because it is the only place, as of October 2026, where a site owner can see the queries an AI actually ran. Every prompt-tracking tool tests prompts its user chose. Grounding queries show the input the retrieval system used instead.

Two research sources point the same way. A July 2026 critical survey of 45 GEO studies (arXiv:2607.14035) found that topical relevance and context position are the most reproducible levers, and that generic heuristics transfer poorly. A 252,000-trial factorial study (arXiv:2605.25517) found that topical relevance and list position are the biggest drivers of being cited first.

Read both with care. The survey is single-author and not peer reviewed. The factorial used a synthetic corpus with two candidate documents per trial. Both measured pages already inside the model's context, so they describe what happens after retrieval, not how a page gets retrieved.

The practical point still holds: topical fit is not keyword coverage. Google's AI optimization guide says you don't have to worry about missing long-tail keywords or every variation of how someone might search. Grounding queries you never targeted are a symptom of that. The system decided your page fits a topic, regardless of the exact phrase.

Grounding query data is useful, but it has limits you should know before acting on it.

• It is a sample. Microsoft scopes its sampling note to the grounding queries metric. You are seeing part of the activity, not all of it. More on that in whether Bing's AI Performance data is sampled.
• Prompt trackers see even less. In Otterly's 2026 experiment, Microsoft's report credited one test site with 42,817 citations, while Otterly's own prompt-based tracker logged 3,934. The time windows may not match, so it is not an exact ratio. It does show how much a prompt sample misses.
• It covers Microsoft surfaces only. In the same experiment, ranking in Bing produced effectively zero ChatGPT citations. Bing data tells you nothing reliable about ChatGPT, Gemini or Google AI Overviews.
• Citation Share is not a rank. Microsoft calls it an observational metric, not a ranking system or a competitive scoreboard.
• It only shows pages already cited. It cannot tell you why another page was never retrieved.

For how Copilot draws on Bing's index in the first place, see how Microsoft Copilot grounds answers in Bing.

Use grounding queries to learn what people ask your pages about, then make those answers clear and consistent. Here is a workable routine.

1. Export and group by page. Pull the grounding queries and map each one to the URL it cited.
2. List the facts each query asks for. A price, a limit, a location, a compatibility detail. These are the facts the AI came to your page to find.
3. State each fact in one plain sentence. A specific fact on a fetchable page gives the engine a sentence it can quote. That is the only positive experimental signal for on-page content. It is not a guarantee of citation.
4. Put it in the HTML before JavaScript runs. Many AI crawlers fetch but do not execute JavaScript, so content that appears only after rendering may be invisible to them.
5. Fix contradictions across your site. If the pricing page, the FAQ and the footer state different numbers, the AI may repeat the wrong one.
6. Don't mass-produce pages for every query. In Otterly's test, two sites with roughly 1,000 AI-generated posts each were removed from Google's index without warning. Copilot citations did not make up for that loss.

What you control is access, clarity and consistency. What Copilot retrieves and cites for a given question stays with Microsoft. If you haven't set up the tool yet, start with whether Bing Webmaster Tools is worth setting up for AI search.

Are grounding queries the same as what users typed into Copilot? No. Grounding queries are the phrases the AI used to retrieve content, not the user's original prompt. AI search systems commonly rewrite a question into one or more search queries, sometimes across several subtopics. One user question can produce several grounding queries, and the wording can differ a lot from what the person typed. That is why the report often shows phrases you never wrote for.

Does a grounding query mean my page ranks for that phrase? No. A grounding query shows that the AI used the phrase to retrieve content it later cited. It is not a position or a rank. Microsoft describes Citation Share as an observational metric, not a ranking system. No AI platform publishes position data inside generative answers, so any tool reporting an AI rank has nothing to check that number against.

Can I see grounding queries for ChatGPT or Google AI Overviews? Not as of October 2026. OpenAI says ChatGPT rewrites queries before searching, and Google describes query fan-out for AI Mode, but neither shows site owners those generated queries. Google Search Console's generative AI report shows impressions only. Bing Webmaster Tools is the only first-party source that exposes the queries an AI actually ran, and it covers Microsoft surfaces.

Should I create new pages to target the grounding queries I see? Usually not. A grounding query you never targeted means an existing page already fits that topic. The better move is to make the relevant fact clear on that page. Google's guide says you don't need to cover every long-tail variation. Publishing many thin pages risks Google's scaled content abuse policy, which removed both test sites in Otterly's 2026 experiment.