Resources · AI Citations and Sources

How Do Citations in Bing Copilot Differ From Bing Search Rankings?

Copilot draws on Bing's index, but it does not simply cite Bing's top results for the question a person typed. It turns the question into shorter grounding queries, retrieves passages for each, and cites the pages that supplied what it used. In Ahrefs' 2025 study of 15,000 questions, only about 17% of the URLs Copilot cited ranked in Bing's top 10 for the same question, though that was the highest overlap of the assistants tested. So a page can rank well and never be cited, or be cited often while ranking modestly for the original question. Bing Webmaster Tools now shows citations and grounding queries separately from search performance, which makes the difference visible.

Why don't citations follow rankings?

Bing search rankings Copilot citations
Query What the person typed Grounding queries Copilot writes, often several per question
Unit Whole pages Passages that answer a specific part of the question
Goal Order the best pages for the query Support the sentences in a generated answer
Count Ten blue links per page of results A handful of sources per answer
Measured in Bing search performance: impressions, clicks, position Bing AI Performance: citations, cited pages, grounding queries

Microsoft describes grounding queries as the key phrases the AI used when retrieving content cited in its answers. Analysts who studied the report describe them as shorter, topical queries generated in the background, not what people type. A page that answers one of those sub-questions precisely can be cited even if it is not in the top ten for the full question.

Which pages does Microsoft Copilot cite most often?

Microsoft has not published a breakdown, but Ahrefs publishes a monthly list of the 50 websites Copilot cites most often, from its Brand Radar data. In the July 2026 edition, Amazon alone accounted for about 14.6% of those citations. The lists name domains, not page types. What the evidence suggests about the pages themselves:

  • Pages that answer one sub-question directly. Because retrieval works passage by passage, a clear section that states a price, a definition, a comparison or a step is easier to cite than a long page that circles the answer.
  • Pages ranking reasonably well in Bing. Copilot still depends on Bing's index, and in Ahrefs' study it overlapped with Bing's top 10 more than ChatGPT, Gemini or Perplexity did.
  • Comparison and list pages for choice questions, which answer several parts of a question at once, as covered in why AI cites comparison pages instead of product pages.
  • Structured sources for local and product questions: Bing Places for businesses, and Microsoft Merchant Center feeds, which Microsoft says help inform organic Copilot results.
  • Fewer social posts than elsewhere. Profound's 2026 analysis of billions of citations found social platforms made up about 4.3% of Copilot citations, compared with about 15% in Google's AI Overviews.

Your own data is the best guide. The Pages view in Bing's AI Performance report lists which of your URLs are cited and how often.

What does Bing's AI Performance report show?

Bing Webmaster Tools added the report in public preview in February 2026. It covers Microsoft Copilot, AI-generated summaries in Bing and select partner integrations, which Microsoft has not named. It shows:

  • Total citations and average cited pages per day.
  • Grounding queries linked to your cited pages, based on a sample of citation activity.
  • Page-level citation counts.
  • Since June 2026, intents for grounding queries, such as informational, topics, your citation share, and a compare view.

Microsoft is explicit that citation counts show how often pages are cited, not their importance, ranking or placement within an answer, and the report has no click data.

Does this matter beyond Copilot?

Probably. In February 2026 Microsoft wrote that its grounding technology “powers nearly every major AI assistant in the market”, and the AI Performance report already counts citations from unnamed partner integrations as well as Copilot and Bing. Microsoft has not said which assistants those are, so treat the report as a partial view of how Bing-grounded answers use your pages rather than a Copilot-only figure. It also means the gap between ranking and citation is not a Copilot quirk: any assistant that writes its own sub-queries and retrieves passages will cite differently from a ranked list of pages. Improving how clearly your pages answer specific sub-questions helps wherever that pattern applies.

How can you compare the two?

  1. Export your top pages by Bing clicks from search performance, and your top pages by citations from AI Performance.
  2. Look for pages cited but rarely clicked. They answer something Copilot needs; consider whether they lead visitors onward.
  3. Look for pages that rank but are rarely cited. Check whether the answer is buried, vague or split across sections.
  4. Read the grounding queries for your most-cited pages and treat them as topics your content answers well.
  5. Check new pages get indexed quickly, since Copilot can only cite what Bing has, as covered in how Bing, Copilot and IndexNow work together.

A wider audit of where AI answers cite you is described in what a citation audit is.

Why can citations be high while traffic stays low?

Citations are not visits. A Copilot answer may use your page without the reader clicking, and some grounding activity is triggered by automated tools rather than people. One agency reported more than 1,000 Bing AI citations for an article that had almost no impressions in traditional search. Treat citations as a sign your content shapes answers, and measure business results separately.

How should you optimise for both?

  • Keep Bing SEO sound: Bing Webmaster Tools verified, sitemap submitted, IndexNow on, no crawl blocks.
  • Write answer-first sections under question-style headings, each able to stand alone.
  • State facts plainly: prices, dates, figures and names in text, not only in images or scripts.
  • Keep Bing Places and Merchant Center current for local and product questions, as covered in why Copilot cites Bing Places but not your location page.
  • Earn coverage on sites Copilot already cites for your topic.

Bing's wider role across assistants is covered in why Bing matters for AI search visibility.

We have not run our own study of Copilot citations. This page is based on Microsoft's Bing Webmaster documentation and blog posts, and published studies by Ahrefs and Profound.

What might this look like in practice?

Consider an illustrative HR software company, Shiftmarrow, whose “shift scheduling software” page ranked third in Bing but was rarely cited. Bing's AI Performance report showed its most-cited page was a short guide to overtime rules, cited for grounding queries about overtime calculation. The team added a clear section to the product page answering “how does shift scheduling software handle overtime”, with an example, and linked it from the guide. Over the next two months, citations of the product page rose steadily while its Bing ranking stayed the same.

What do people ask about Copilot citations and Bing rankings?

Does ranking first in Bing guarantee a Copilot citation?

No. Copilot cites passages that answer its grounding queries, which may come from lower-ranked pages.

What are grounding queries?

Short search phrases Copilot generates to retrieve information for an answer. They are not what users typed.

Does the AI Performance report show clicks?

No. It shows citations, cited pages and grounding queries, not clicks or positions.

Can I see which Copilot answers cited me?

Not individually. The report aggregates citations by page and grounding query.

Published by The Kovmere Team · updated Sep 29, 2026 · More resources