How to Track Brand Mentions in Microsoft Copilot in 2026

Microsoft Copilot has no analytics. Here is the agency method for tracking brand mentions in Copilot with two booleans - named and cited - backed by stored raw answers.

TL;DR: Microsoft Copilot ships no analytics, so tracking brand mentions means running your own prompt set through Copilot, reading each answer, and recording two separate booleans - was the brand named in the prose, and was its domain among the numbered footnote sources. Copilot is Bing-grounded and reaches buyers through Windows, Edge, and Microsoft 365, so for enterprise clients it is the engine you cannot skip, even though its consumer share trails ChatGPT. Store the raw answer behind every number and diff against the brand's own history. Never a single "visibility score."

Copilot is easy to underrate. Its consumer traffic is smaller than ChatGPT's, so agencies chasing raw reach put it last or drop it entirely. But Copilot rides Microsoft's distribution instead of competing for it: it sits on the Windows taskbar, in the Edge sidebar, and inside Word, Excel, Outlook, and Teams through Microsoft 365. When a B2B buyer or an IT decision-maker asks "which vendor should we shortlist," they are often asking Copilot without ever opening a fresh browser tab. For enterprise clients, that is exactly the audience that signs the contract. If your client is missing from that answer - or named while Copilot points the reader at a competitor - the deal leaks upstream, invisibly. So it gets tracked. The problem is the same as every other answer engine: Copilot has no Search Console, no impressions, no export. You build the measurement yourself.

Why a single Copilot "score" is the wrong tool

Most trackers will hand you one number: "62% visible on Copilot." It feels like a status and tells you nothing you can act on. Sixty-two percent of what - named, cited, or both, blurred together? You cannot defend it to a client, and you cannot check it, because there is no artifact underneath. A merged score also hides the failure mode that actually loses enterprise deals: Copilot naming your client while every footnote points at a competitor's comparison page. The number can drift upward while the reader keeps clicking through to someone else. The fix is to stop merging signals and read two facts off each answer instead.

The two booleans, read off a Copilot answer

Every Copilot answer gives you two independent facts. Keep them apart - see mention vs citation for why they never collapse into one:

• Mention - is the brand named in the prose Copilot writes back? Yes or no.
• Citation - is the brand's own domain among the numbered footnote sources under that answer? Yes or no.

Copilot makes both unusually easy to read because of how it presents an answer. It writes conversational prose with superscript footnote markers inline, and each marker maps to a numbered source card in a list beneath the answer. The prose is where you read the mention. The numbered list is where you read the citation.

How Copilot's citations work, and why Bing matters

Copilot is Bing-grounded: its retrieval comes from Bing's web index, not Google's. That has one practical consequence for the citation boolean. A domain can only be cited if Bing can find it, so a client page that ranks well in Google but is thin or unindexed in Bing may never surface as a Copilot source, no matter how strong it looks elsewhere. This is a real, checkable input - and a boundary you should say out loud. Bing indexing makes a domain eligible to be cited; it does not make Copilot cite it, and it does not make Copilot name the brand. Nobody controls what the engine says. You measure what it said.

One more thing to record as metadata: which Copilot you ran. The public, web-grounded Copilot at copilot.microsoft.com and the Edge sidebar answer the open-web question this method audits. Microsoft 365 Copilot in "work" mode grounds partly on a tenant's own documents - a different surface, a different answer, and not what an external GEO audit reads. Note the surface on every run.

Signal	Where to read it on a Copilot answer	Boolean
Mention	The conversational prose Copilot writes - is the brand named in the sentence the reader sees?	Named = Y / N
Citation	The numbered footnote markers and the source list beneath - is the brand's own domain one of them?	Own domain cited = Y / N

The method

1. Build a prompt set. 20-50 real buyer questions in the client's category - what an enterprise prospect or IT lead actually asks, not brand-name lookups. Question-shaped queries surface the truth; brand lookups flatter everyone.
2. Run each prompt in Copilot deliberately. Fresh session. Record the surface (web Copilot vs Edge sidebar), whether web search was engaged, and the country - these change the answer, so they are part of the run, not trivia.
3. Capture the verbatim answer and its numbered sources. Copy the exact prose and the full list of cited URLs behind the footnotes. Not a summary - the artifact.
4. Mark the two booleans: named in prose (Y/N), own domain in the footnotes (Y/N).
5. Store the raw answer behind the number. This is the step trackers skip and the step that makes the report survive scrutiny. A month later, "you moved out of the source gap on this prompt" only means something if you can reopen both answers.
6. Diff against the client's own history, on a schedule. The baseline is their past Copilot runs, not an invented target. Flag a change only when it moves beyond run-to-run noise.

A quick worked example makes it concrete. Say the client sells an ITSM platform and the buyer question is "best IT service management tool for a mid-size company." You run it in web Copilot, US, and capture the prose plus the numbered sources. Copilot names four vendors and lists six footnotes. Now read the booleans: if the client is named in the prose and one footnote links their domain, that prompt is a stronghold - defend it. If Copilot names the client but every footnote points at a review directory and a competitor comparison, that is a source gap: the reader hears the name and clicks through to someone else. If Copilot never says the client's name but footnote four is one of their own blog posts, that is a conversion gap: the content did the work and the brand vanished. Same prompt, same engine, opposite fixes - which is exactly why a single "Copilot visibility %" is useless and the two booleans are not. (This is illustrative; the actual answer is whatever Copilot returns on the day you run it.)

Why enterprise agencies should not skip Copilot

If your client sells to businesses, Copilot's smaller consumer share is the wrong metric to judge it by. The right metric is where the buyer sits. Microsoft put Copilot on the taskbar of the operating system those buyers already run, in the browser their IT department standardized on, and inside the productivity suite where they draft the shortlist. That is distribution ChatGPT does not have inside the enterprise. Auditing ChatGPT and Perplexity while ignoring Copilot leaves a hole exactly where B2B purchase decisions get made. Jincove reads these same two booleans off Copilot alongside the other engines - see the features for how each answer is stored and diffed.

Track all six engines, not just Copilot

Copilot is one surface. Buyers also land in ChatGPT, Perplexity, Gemini, Google AI Overviews, and Google AI Mode, and the same brand can be a stronghold in one and whitespace in another. The reading discipline is identical everywhere - the same two booleans, the same stored answer - which is why the method transfers cleanly. Start with the ChatGPT tracking method for the sibling walkthrough, then run the same prompt set across all six and sort every answer into a quadrant with the source gap vs conversion gap framework. Two booleans, four boxes, one fixed move each.

One practical note on where to begin. The free audit hand-runs three engines - ChatGPT, Perplexity, and Gemini - so it does not cover Copilot; Copilot is the sixth engine, included in the paid six-engine monitoring. That is the right split for enterprise clients: prove the method on the free three first, then bring Copilot in where the B2B buyers actually are.

See it on your own brand first

Request a free GEO audit: send one URL and an email, and we hand-run ChatGPT, Perplexity, and Gemini, then reply with the exact answers, the cited sources, and whether each named or only cited the brand. No card, no account. It is the fastest way to see the two-boolean method on a brand you care about - and the natural on-ramp to monitoring all six engines, Copilot included.