How do you get cited by Microsoft Copilot?
Copilot grounds its web answers in the Bing index, so Bing decides what Copilot is able to cite and your Bing hygiene is the first lever. After that, selection works like every other engine: a heading that matches the question, a first passage that answers it on its own, and a named source beside every claim. Copilot is also the one assistant that shows you its work. The AI Performance report in Bing Webmaster Tools, launched in beta on 10 February 2026 and expanded that June, names the URLs Copilot cited and the grounding queries that produced them, which no other engine publishes. The commercial reason to care is where Copilot sits rather than how big Bing is: Microsoft reported Microsoft 365 Copilot passing 30 million paid seats in July 2026, which is an assistant embedded in the tools people use at work.
Where does Microsoft Copilot get its sources from?
Copilot gets its web sources from the Bing index. Bingbot crawls and renders pages, builds the index that serves Bing Search, and that same index feeds Copilot and Microsoft's grounding APIs. Microsoft's own documentation for web search in Copilot Studio describes the flow: the assistant sends generated search queries to the Bing service and composes an answer from what comes back, with citations.
The queries Copilot sends are not the words the user typed. Microsoft calls them grounding queries: internally generated searches written to retrieve what the model needs, in the same family as the fan-out behaviour every answer engine uses. One buyer question can produce several of them, and each one runs its own retrieval against Bing before the answer is written.
The dependency chain is short and unforgiving. If bingbot cannot fetch a page, Bing cannot index it. If Bing cannot index it, the page is not in the candidate set for any grounding query. If Bing cannot decide which of three near-identical URLs is canonical, its confidence in all three drops and none of them gets picked. Every problem that hurts you in Bing hurts you twice, once in search and once in Copilot.
Bing Webmaster Tools is the place to verify all of that, and it is free. Confirm the pages you care about are indexed, submit your sitemap, check that URL inspection renders the content and not an empty shell, and use IndexNow to push updates rather than waiting for a recrawl. Our guide to checking whether AI can read your website covers the crawlability side in order of impact.
Why does Copilot matter if Bing's search share is small?
Copilot matters because of where it sits, not because of how much consumer search Bing wins. Copilot is the assistant inside Windows, Edge, Word, Outlook and Teams, which is where a lot of commercial research actually happens, and it arrives in front of people who never chose an AI product at all. Microsoft reported in its FY26 fourth-quarter results in July 2026 that Microsoft 365 Copilot had passed 30 million paid seats, against 15 million reported in January 2026 and 20 million in April.
Bing's search footprint is also concentrated in exactly that context. Statcounter's desktop figures put Bing at 8.84% of worldwide desktop search in August 2026 and at 0.6% on mobile. Bing is a work-machine search engine, which makes it a poor proxy for consumer reach and a good proxy for the desk where a procurement question gets typed.
The asymmetry worth acting on is cost. Competing for Google's AI surfaces means competing with everyone, because everyone optimises for Google. Bing receives a fraction of the same attention, so ordinary technical hygiene and one well-answered page can earn a citation that would take far more work on the Google side. For a business selling into enterprises running Microsoft 365, that is the cheapest AI visibility available.
Treat Copilot as its own engine rather than a Bing by-product. Presence in one assistant tells you little about presence in another, which our note on why a business shows up in ChatGPT but not Google AI Overviews works through with the published overlap data.
How do you check whether Copilot is citing you?
Check Copilot citations in the AI Performance report inside Bing Webmaster Tools, which is the only first-party citation report any answer engine publishes. Search Engine Land reported the beta launch on 10 February 2026. The report shows which of your URLs were cited in Copilot and Bing's AI answers, which queries produced the citation, and how the volume moves week to week.
The grounding queries are the part worth reading closely. Because they are the searches Copilot wrote for itself, not the ones users typed, they show you the question behind the question: the sub-queries your page was considered for, in the engine's own phrasing. That is a direct read on vocabulary alignment, and it is information no keyword tool can give you.
Microsoft expanded the report in June 2026. The Bing blog's announcement of the new AI visibility insights added four views: Intents, Topics, Citation Share and Compare. Citation Share is the commercially useful one, because it shows what proportion of the citations for a grounding query went to you across the full set of cited sources rather than leaving you to guess. Compare overlays an earlier period on the current one, so you can see whether that share is moving.
Read the report with one caveat in mind. Microsoft has not shipped click metrics yet, so a citation confirms your page was used as a source and says nothing about whether anyone clicked through. Treat the count as a visibility measure rather than a traffic measure, and pair it with the sampling approach in our guide to tracking AI visibility across engines so that one engine's reporting does not become your whole picture.
Does Copilot cite different sources from ChatGPT and Gemini?
Copilot does cite differently, and the difference runs in two directions. Each assistant retrieves from its own index, so the candidate sets barely overlap, and Copilot's attribution behaviour also measures differently from its rivals. The EBU and BBC study of news integrity in AI assistants, published in October 2025, had journalists from 22 public service media organisations in 18 countries review more than 3,000 responses in 14 languages.
On sourcing, Copilot came out at the top of that group. The study found significant sourcing problems in 15% of Copilot responses, the same as Perplexity, against 24% for ChatGPT and 72% for Gemini. Across all four assistants, 45% of responses carried at least one significant issue and 31% had sourcing problems, so Copilot leading this field is a relative result rather than a clean bill of health.
The practical meaning for a site owner is that a Copilot citation is comparatively literal. An assistant that attributes carefully is an assistant that needs a passage it can point at, which rewards pages where the answer sits in one place instead of accumulating across four paragraphs. Writing for extraction pays off faster here than on engines that paraphrase loosely across many sources.
The narrower index cuts the other way too. Fewer candidate pages means fewer slots, and a source that does not exist in Bing simply is not considered, however strong it is elsewhere. Copilot is the engine where a technical indexing failure is most likely to be the entire explanation for absence.
What should you fix first to get cited by Copilot?
Fix the Bing side first, because nothing else can compensate for not being in the index Copilot reads. Then work down through extraction, evidence and third-party presence, checking each against the AI Performance report and not against a general AI visibility dashboard.
| Fix | Why it matters for Copilot | Where to verify it |
|---|---|---|
| Let bingbot crawl and render | The Bing index is Copilot's candidate set, and blocked or client-rendered pages never enter it | Bing Webmaster Tools URL inspection |
| One canonical URL per page | Duplicate URLs split signals and lower Bing's confidence in picking any of them for grounding | Bing Webmaster Tools site scan |
| Push updates with IndexNow | Copilot answers from the indexed version, so a stale index shows a stale answer | IndexNow submission log in Bing Webmaster Tools |
| Answer-first passages under question headings | Copilot cites the passage it quotes, and careful attribution rewards a single quotable answer | AI Performance report, cited URLs |
| Named sources beside every number | Grounded claims survive the check a model makes before attributing a sentence to you | Your own page audit |
| Presence on third-party pages | Buying questions retrieve comparisons and reviews rather than vendor sites | Citation Share view, added June 2026 |
Work the list against grounding queries rather than against your own keyword list. If the report shows Copilot retrieving for a phrasing you never targeted, the cheapest available win is a heading in that exact phrasing on the page that already ranks. Our AI citation checklist sets out the page-level items in order of measured impact, and how AI engines choose which sources to cite covers the selection mechanism these fixes are aimed at.
What else do people ask about Copilot citations?
How do you get cited by Microsoft Copilot?
Get into the Bing index cleanly, then make each page answer one buyer question outright in its opening sentences. Copilot grounds its web answers by sending its own generated search queries to Bing, so a page Bing cannot crawl, index or confidently pick a canonical URL for is a page Copilot cannot cite. Once you are indexed, the selection is about extractability: a heading that matches the question and a first passage that answers it without needing the rest of the page.
Where does Microsoft Copilot get its sources from?
Copilot gets its web sources from the Bing index, fetched by bingbot, which also supplies Bing Search and Microsoft's grounding APIs. When a question needs live information, Copilot writes its own search queries, retrieves candidate pages from Bing, and cites the passages it uses. Microsoft calls those internally generated searches grounding queries, and reports them to site owners in Bing Webmaster Tools.
Why does Copilot matter if Bing's search share is small?
Because Copilot reaches people at work, where the buying research happens, rather than competing for consumer search share. Microsoft reported in its FY26 fourth-quarter results in July 2026 that Microsoft 365 Copilot had passed 30 million paid seats, up from 15 million reported in January 2026. Bing's own footprint is also concentrated where that work happens: according to Statcounter, Bing held 8.84% of worldwide desktop search in August 2026 against 0.6% on mobile.
How do you check whether Copilot is citing you?
Use the AI Performance report in Bing Webmaster Tools, which Microsoft launched in beta on 10 February 2026. It shows which of your URLs are cited in Copilot and Bing's AI answers, the grounding queries that triggered those citations, and how citation activity moves over time. Microsoft added Intents, Topics, Citation Share and Compare views in June 2026. There is no click data yet, so a citation confirms your page was used as a source, not that anyone visited.
Does Copilot cite different sources from ChatGPT and Gemini?
Yes, because each assistant retrieves from a different index, and Copilot's sourcing behaviour also measures differently. In the EBU and BBC study of news answers published in October 2025, which reviewed more than 3,000 responses in 14 languages, Copilot had significant sourcing problems in 15% of responses, against 24% for ChatGPT and 72% for Gemini. Copilot attributes more carefully and retrieves from a narrower index, so presence in Bing matters more for Copilot than for any other assistant.