Copilot Brand Tracking: Mentions, Citations and Grounding Queries

Microsoft launched the AI Performance report in Bing Webmaster Tools as a public preview on 10 February 2026 and expanded it on 16 June 2026. Its stated coverage is "Microsoft Copilot, AI-generated summaries in Bing, and select partner integrations."

As of October 2026, the report exposes:

• Total Citations: how often your pages were cited in AI answers.
• Average Cited Pages: how many of your pages get cited.
• Page-level citation activity: which URLs earn the citations.
• Grounding queries: the phrases the AI used to retrieve content it then cited.
• Intents and Topics: groupings such as Informational, Commercial, Navigational, Learn and Solve, Research, Creation and Local.
• Citation Share and Compare: your share of citations and period comparisons.

This is first-party data from the operator of the engine. No third-party tool has access to the same counts. If you track Copilot at all, start here. Our walkthrough on reading Bing's AI Performance report covers each panel.

One limit matters for brand work. Every metric Microsoft lists counts citations. None of them describes what Copilot actually said about you. A citation can sit next to a wrong price or an outdated product name, which is why fixing wrong facts in Copilot answers is a separate job.

Microsoft defines grounding queries as "the key phrases the AI used when retrieving content that was referenced in AI-generated answers." That is the input to retrieval, not the user's prompt.

The distinction is the whole point. AI engines do not search for what users type. OpenAI says ChatGPT search "rewrites your query into one or more targeted queries" before sending them to search providers. Google's Gemini API grounding "automatically generates one or multiple search queries." Copilot grounds its answers in Bing through a similar multi-step pipeline.

So a prompt tracker tests the questions you chose. Grounding queries show the searches the engine actually ran. As of October 2026, Bing is the only vendor that exposes them.

What to do with them:

1. Export the grounding queries for your top cited pages. The steps are in exporting and analyzing Bing grounding queries.
2. Compare them with the prompts your tracker runs. Nobody has published how much the two overlap, so your own comparison is the only answer you will get. See comparing grounding queries with your prompt set.
3. Rewrite your prompt set toward the phrasing the engine uses.

One caveat. Microsoft's note that "the data shown represents a sample of overall citation activity" sits inside the grounding queries subsection. It applies to that metric, not the whole report. More detail is in whether Bing's AI Performance data is sampled.

Citation Share invites misreading. Microsoft describes it as "an observational metric" and says it is "not a ranking system or a competitive scoreboard." It also says the metric "does not expose competitor domains, represent traffic share, or assign quality scores to content."

Take that framing literally. Citation Share is a share, and its denominator is dominated by whoever already wins the query. If a large competitor gains citations on the same grounding queries, your share can fall while your absolute citations rise.

How to report it:

• Lead with Total Citations and page-level counts. Those are your numbers.
• Show Citation Share as context, next to the absolute count for the same period.
• Never present it as a rank or a position. No engine publishes position data inside AI answers to check such a number against.

OtterlyAI ran a public experiment in 2026: two new domains, each loaded with roughly 1,000 AI-generated posts. Google deindexed both. Bing kept them. Copilot kept citing them.

The tracking numbers are the useful part here. Otterly's prompt-based tracker logged 3,934 citations for capybara-world.com across seven engines, and 2,718 of those (69%) came from Microsoft Copilot. Microsoft's own AI Performance report credited the same site with 42,817 citations, more than ten times Otterly's total.

Read that carefully. The time windows may not match, so it is not an exact ratio. It still shows what any prompt tracker is: a sample of answers to the prompts someone chose to ask. Where an engine publishes its own count, that count sees more.

The same experiment has a second lesson. Ranking in Bing produced effectively zero ChatGPT citations, even with ChatGPT's crawlers visiting the sites. Copilot data tells you about Copilot. It is not a proxy for other engines.

The API gap is real. The published Bing Webmaster API reference was last updated in November 2023 and contains no citation, grounding, Copilot, intent or topic endpoints. Microsoft has indicated API access may arrive during 2026, so check the current state in whether there is an API for the AI Performance report. Until then, any tool showing this data got it from a manual export.

Tracking only works if Copilot can reach and index you. Fix the gates first, then measure.

1. Check crawler access. Bing's index backs Copilot, so bingbot must be allowed in robots.txt and not blocked by your CDN or firewall. A robots.txt allow means nothing if the firewall returns 403. The AI Crawler Checker requests a page as each crawler and shows which ones get turned away.
2. Confirm Bing index presence. Verify the site in Bing Webmaster Tools. Google's index status tells you nothing about Bing's, as the Otterly test showed.
3. Check for accidental suppression. Bing says content marked data-nosnippet "is still indexed normally, but it will be excluded from snippets and AI summaries." A template-level nosnippet or noindex does the same at page level.
4. Export the AI Performance report monthly. Keep Total Citations, top cited pages and grounding queries in one sheet.
5. Pair it with answer checks. Run a small set of buyer questions to see what Copilot actually says, rebuilt from your grounding queries.

What stays out of your control: whether Copilot retrieves you for a given query, and which retrieved sources it cites. Access, indexing and correct facts on your own pages are what you can fix.

Does the AI Performance report cover ChatGPT? No. Microsoft scopes it to Microsoft Copilot, AI-generated summaries in Bing, and select partner integrations. It does not name ChatGPT. In Otterly's 2026 experiment, sites that ranked in Bing and earned thousands of Copilot citations got effectively zero ChatGPT citations. Track ChatGPT separately, and check that OAI-SearchBot, which OpenAI documents as the crawler for ChatGPT search, can reach your site.

Can the report show how often competitors are cited? Not by name. Microsoft states that Citation Share does not expose competitor domains. You see your own share of citations, not who holds the rest. To compare against competitors, you need answer-level checks on shared buyer questions, and you should compare brands of similar size. Brand prominence is strongly associated with citation volume, so comparing a small brand with a giant mostly measures size.

Is grounding query data complete? No. Microsoft says the data in that subsection represents a sample of overall citation activity. The sampling note sits inside the grounding queries section, so do not assume it applies to Total Citations or page-level counts, which carry different caveats. Treat grounding queries as a reliable indicator of phrasing and topics, not as a full log of every search Copilot ran.

Why does my Copilot citation count differ from my tracking tool? Your tool counts answers to the prompts it asks. Microsoft counts citations across real usage. In Otterly's experiment, Microsoft's report showed 42,817 citations for one site while a prompt tracker logged 3,934 across all engines. Time windows may differ, so the gap is not an exact ratio, but the first-party number will usually see more because it is not limited to a chosen prompt list.

Do more Copilot citations mean more traffic? Not necessarily, and the report does not claim they do. Microsoft states Citation Share does not represent traffic share. A citation shows your page was referenced in an answer, not that anyone clicked. Measure referral visits separately in your analytics, and report citations and traffic as two different numbers rather than treating one as a stand-in for the other.