Microsoft Copilot visibility: how Copilot finds websites, and how to improve your odds

Microsoft Copilot answers questions by combining its underlying language models with live grounding from Bing's search index. That one architectural fact drives everything about Copilot visibility: if Bing cannot find and rank you, Copilot has very little to cite. Here is how the grounding works, how to check your own visibility today, and the improvements that move it.

How Microsoft Copilot decides what to cite

Microsoft Copilot is grounded in Bing. When Copilot needs current information, it runs a web search against Bing’s index, reads the retrieved pages, and composes an answer with citations to a small number of sources. The model supplies the language; Bing supplies the candidate sources. So a brand’s Copilot visibility rests on two layers:

The retrieval layer (Bing). Can Bing find, index and rank your pages for the question being asked? If you are thinly indexed in Bing, Copilot’s grounding search returns competitors instead of you, and no amount of content quality on your side changes the answer.
The selection layer (the model). Of the pages Bing returns, which does Copilot actually quote? This is where citable structure matters: clear entity signals, answer-first passages the model can lift, and content that directly addresses the question.

If this sounds familiar, it should: ChatGPT’s live search layer also draws largely on Bing’s index. The work you do for Copilot visibility and the work you do for ChatGPT’s live layer overlap substantially, which is why a site that fixes its Bing fundamentals often improves in both.

Why Google rankings do not carry over

The most common Copilot surprise is a site that ranks well on Google and barely exists in Copilot. The explanation is structural, not mysterious: Copilot never consults Google. Backlink authority Google has credited you with, rankings you have earned, none of it is visible to Bing unless Bing has independently indexed and weighted your site. Bing’s index is smaller and its crawler behaves differently, so gaps that never show up in a Google-first SEO programme can quietly erase you from every Bing-grounded surface: Copilot, Bing search, and ChatGPT’s live retrieval.

How to check your Copilot visibility manually

An honest note first: SearchScore’s Tracker does not currently scan Microsoft Copilot. It runs structured weekly queries against six other engines (ChatGPT, Claude, Gemini, DeepSeek, Grok and Perplexity). For Copilot specifically, checking is a manual job today. Here is the method we recommend:

Write down the five to ten questions your customers actually ask. Decision-stage questions (“best [your category] for [use case]”, “alternatives to [competitor]”), not brand-named ones. Asking Copilot about your own brand by name only tests whether it knows you exist, not whether it recommends you.
Ask them in Copilot, signed out or in a fresh session. Use copilot.microsoft.com or the Copilot app. A fresh session avoids your own history nudging the answers.
Record three things per question: whether you were named at all, what position you appeared in if the answer was a list, and which competitors were named instead of you.
Check the citations. Copilot links its sources. Note which of your pages (if any) it cited, and which third-party pages carried competitors into the answer. Those third-party pages are your outreach targets.
Cross-check Bing directly. Search the same questions on bing.com. If you are absent from Bing’s first page, that is usually the root cause of being absent from Copilot, and it tells you the fix is retrieval, not content.
Repeat on a schedule. One check is a snapshot. Answers shift as Bing re-crawls and as models update, so put the same question set in a spreadsheet and re-run it monthly.

How to improve Copilot visibility

Because Copilot is Bing-grounded, the levers are mostly Bing levers plus citable structure:

Verify your site in Bing Webmaster Tools. It is the direct channel for finding out what Bing has and has not indexed, and for submitting what is missing.
Adopt IndexNow. Bing consumes IndexNow pings, which get new and updated pages into the index quickly instead of waiting for a re-crawl. SearchScore’s audit checks IndexNow and Bing-verification signals as part of its AI visibility scoring for exactly this reason.
Do not block Bing’s crawler. Check robots.txt for rules that catch bingbot, including well-intentioned blanket “block all bots” rules.
Fix rendering. If your content only exists after JavaScript runs, crawlers may see an empty page. Server-rendered content is reliably retrievable.
Structure pages so a passage can be lifted. Question headings with a direct 40 to 60 word answer underneath give the model a clean quote. Dense prose forces it to paraphrase, and it usually picks a source it can quote instead.
Sharpen entity signals. Organisation schema, a consistent name, address and description across the web, and a clear about page help the model attribute an answer to you confidently.
Earn third-party mentions Bing can see. Copilot frequently cites roundups, directories and review sites rather than vendors themselves. Being present in the pages that already win these answers is often faster than trying to outrank them.

Where SearchScore fits

Two honest boundaries and two genuine helps:

The free SearchScore audit does not query Copilot, but it does score the underlying signals Copilot’s grounding depends on: AI crawler access, Bing-side retrievability signals like IndexNow, structured data, entity clarity and citable structure. Fixing those moves every Bing-grounded surface at once.
The SearchScore Tracker covers six engines, not Copilot (yet). It runs your customers’ questions against ChatGPT, Claude, Gemini, DeepSeek, Grok and Perplexity every week and reports who got cited, at what position, with what sentiment. Because ChatGPT’s live layer shares Copilot’s Bing grounding, ChatGPT citation results are the closest tracked proxy for Copilot performance available today.

Frequently asked questions

Does Microsoft Copilot use Google or Bing?

Bing. Copilot grounds its web answers by searching Bing’s index, retrieving pages, and citing a handful of them. Google rankings have no direct influence on Copilot answers, which is why a site can lead on Google and still be invisible in Copilot if Bing has indexed it poorly.

How do I check if Copilot recommends my business?

Ask Copilot the decision-stage questions your customers ask, in a fresh session, and record whether you are named, your position in any list, and who is cited instead. Cross-check the same queries on bing.com: absence from Bing’s first page usually explains absence from Copilot. Re-run the same questions monthly, because answers shift as Bing re-crawls.

Does SearchScore track Microsoft Copilot?

Not yet. The SearchScore Tracker scans six engines weekly: ChatGPT, Claude, Gemini, DeepSeek, Grok and Perplexity. For Copilot, checking is manual today. The audit does score Bing-side signals such as IndexNow and crawler access that Copilot’s grounding depends on, and ChatGPT’s Bing-backed live layer is the closest tracked proxy.

Is Copilot visibility the same as Bing SEO?

Largely, plus one extra layer. Bing SEO gets you into the candidate set Copilot retrieves from; citable structure and clear entity signals decide whether the model actually quotes you from that set. You need both: retrievable in Bing, and quotable once retrieved.