Bing AI Performance Report: How to Track Copilot Citations

What is the Bing AI Performance report?

The Bing AI Performance report is a free dashboard in Bing Webmaster Tools that shows how often your content appears as a cited source in Microsoft Copilot, AI-generated summaries in Bing, and select partner integrations. In short, it counts citations, not clicks. That makes it a first-party view of your visibility inside AI answers.

Microsoft announced the report in February 2026 as a public preview. In the official Bing Webmaster announcement, the company frames it as an early step toward Generative Engine Optimization (GEO) tooling.

We see it as a new window next to your classic ranking reports. However, users now get many answers without clicking a blue link. Until now, you could not see whether your pages fed those answers. So this report starts to close that gap.

It is still a preview. Menus and metrics may change, so treat the steps in this guide as a map, not a fixed manual.

Which AI surfaces does the report cover?

According to Microsoft, the report covers three areas: Microsoft Copilot, AI-generated summaries in Bing, and select partner integrations. In other words, the data comes from the Microsoft ecosystem only.

The announcement does not list ChatGPT, Perplexity, Gemini, or Google AI Overviews. So do not read the numbers as your total AI visibility. Instead, you see Microsoft-side citations and nothing else.

The report also respects robots.txt and other content controls. If you block Bing crawlers, your pages may not appear at all. Therefore, check this before you open the dashboard.

Copilot reaches people through Windows and Edge, so it can matter more than Bing's search share suggests. Also, measuring it costs nothing, which makes the decision easy.

Which metrics does the Bing AI Performance report show?

The report gives you four core data points plus a timeline. They are total citations, average cited pages, grounding query samples, and page-level citation activity. The table below explains what each one tells you.

Microsoft says the grounding query list is a sample of overall activity. Read it as a hint about direction, not as a full keyword export.

How do you read total citations?

Total citations counts how many times your site appeared as a source in AI answers during the date range you pick. However, it is not a ranking or a traffic number. Microsoft states that the metrics reflect citation frequency, not page importance, ranking, or placement.

So do not attach absolute meaning to the figure. Instead, the trend matters more. If the last 30 days beat the previous 30, your content is working harder in the AI layer. Then, if it drops, look for the cause.

• Compare two full periods instead of single days.
• Match citation changes to your publishing and update calendar.
• Keep citations apart from organic clicks, because they measure different things.

In short, total citations works like a thermometer. It will not name the illness, but it tells you when the temperature changes.

What do average cited pages tell you?

This metric shows how many unique URLs from your site appeared as sources per day, on average. However, total citations can pile up on a few pages. This number shows how widely the citations spread.

For example, if total citations are high but the page count is low, your visibility rests on a handful of articles. That is risky, because one slip hurts everything. If one of them slips, your whole AI presence slips with it.

On the other hand, a rising page count suggests your content library is gaining trust as a whole. In practice, that often points to deep, well-clustered topic coverage.

On client projects, we use this ratio to decide whether to widen the content plan by topic or to strengthen the pages that already perform.

What is a grounding query and why does it matter?

A grounding query is the search phrase an AI system uses to find sources while it builds an answer. However, it does not have to match what the user typed. The system can split a question into parts and rewrite them.

Therefore, the phrases in your report are valuable. They show which wording and which sub-question made your page useful. Also, classic keyword research rarely gives you this view.

When you open the list, look for these patterns:

• Unexpected phrases, which suggest your page is chosen for a different intent.
• Long, question-style phrases, which you can move into your headings.
• Phrases that include your brand name, which signal brand awareness.

Because the data is a sample, never rebuild a strategy around one phrase. We treat the list as raw material for the content calendar.

What did the June update add to the report?

In June 2026, Microsoft added four features: Intents, Topics, Citation Share, and Compare. According to the Bing search blog, all four are in preview and available worldwide. As a result, the report moved from a simple counter to a topic and intent analysis tool.

Microsoft also notes that citation patterns shift with user behavior, evolving models, freshness signals, and partner refresh cycles.

How do you read Intents and Topics?

Intents shows the purpose behind the queries that cite you. Microsoft lists categories such as informational, commercial, navigational, learn and solve, research, creation, and local. The mix tells you at which stage of the journey your content helps.

For example, suppose most citations are informational. Then your blog content works well. If your service pages earn no commercial citations, though, you have a gap. To fill it, add a clear definition, a price framework, and an FAQ to those pages.

Meanwhile, Topics groups queries into themes. In Microsoft's example, searches about solar panels, solar energy efficiency, and residential installation land in one Solar Energy cluster. During preview, some labels may stay broad, especially in niche fields.

• Compare the cluster labels with your own content map.
• Link cited topics to your pillar pages.
• Plan new content for clusters where you never appear.

In short, reading both views together is best. Intents gives you the why, and Topics gives you the what.

What does Citation Share show, and what does it not show?

Citation Share shows the percentage of citations that went to your site out of all citations shown across all sites for the same grounding query. Microsoft stresses that it is an observational metric, not a ranking system or a competitive scoreboard.

First, the report does not reveal competitor domains. Second, it also does not show your traffic share. So you cannot conclude that a 40 percent share means 40 percent of the traffic.

What is it good for, then? It helps in three ways:

1. You notice new areas of expertise.
2. You find coverage gaps.
3. You see how AI systems classify your content by theme.

We do not set Citation Share as a goal on its own. We read it together with the monthly trend.

How do you access the Bing AI Performance report?

The report lives inside Bing Webmaster Tools. First you need to verify your site. Importing from Google Search Console can speed this up. Once verification finishes, you open the AI Performance section in the dashboard.

1. Sign in to Bing Webmaster Tools at bing.com/webmasters and add your site.
2. Complete verification with an HTML tag, a DNS record, or an import.
3. Open the AI Performance section in the left menu.
4. Pick a date range and review total citations and the page table.
5. Browse the Intents, Topics, and Compare views in turn.

Because the report is in preview, menu names or layout may change. Adapt these steps to what you see on screen.

A free Microsoft account is enough. Still, setup rarely takes more than half an hour. Data may take a while to build, so check again after a week or two if the first view looks empty.

How does it differ from the Search Console AI report?

The two reports measure different ecosystems. Bing's report counts citations in Copilot and Bing summaries. However, Google's reporting depends on Google's own surfaces. Blending them into one total would mislead you.

For the Google side, read our guide to the Search Console AI report. For the full measurement picture, see how to measure AI visibility.

What are the limits of the report?

Remember the biggest limit first: the data is a sample of citation activity, not the full set. The second limit is that the metrics do not show ranking or placement. A page with many citations may not sit at the top of the answer.

• Only Microsoft surfaces are measured, so other assistants stay outside.
• The report is in preview, so metrics and layout can change.
• A citation is not a visit or a sale, because it measures presence only.
• Citation Share shows neither competitor names nor traffic.

Because of these limits, never make the report your only decision tool. Cross-check it with analytics, Search Console, and sales records.

For example, if citations rise while branded searches rise too, the effect may be real. If citations rise and nothing else moves, stay patient and keep measuring.

What actions can you take from the data?

A good report produces a to-do list, not a scorecard. For example, each month we start with three questions. First, which pages earn citations? Second, which topics sit empty? Finally, which indexed pages never appear?

Based on the answers, you can sort actions into three groups:

• Protect: keep cited pages fresh and avoid breaking them.
• Strengthen: add proof, tables, and FAQs to borderline pages.
• Create: plan new, deep pages for topics where you do not appear.

You can also mine the grounding list for headline ideas. Question-style phrases often turn into H2 headings directly. This approach sits at the core of our AI SEO and GEO work.

How do you strengthen pages that already earn citations?

Microsoft's guidance highlights five areas: depth and expertise, structure and clarity, evidence-based content, freshness and accuracy, and format consistency. Together, they also tell you what to touch on cited pages.

• Depth: answer the sub-questions instead of skimming the topic.
• Structure: use headings, tables, and FAQ sections consistently.
• Evidence: add examples, data, and sources, and name the source in the sentence.
• Freshness: update dated facts on a schedule.
• Consistency: make text, images, and video say the same thing.

Paragraphs that answer directly in the first sentence are easier for AI to quote. We explain this in our guide on writing content for AI Overviews. The same logic applies to Copilot.

What should you do about pages with no citations?

First, confirm that Bing indexes the page. A page that is not indexed cannot earn a citation. Next, review your robots.txt rules. For instance, needless blocks can cut crawlers off from your content.

After that, look at the content itself. Does the page truly answer the question? Are the headings clear? Do competing pages go deeper? In practice, the problem is thin content, not a technical fault.

1. Check the index status in Bing Webmaster Tools.
2. Review robots.txt and meta robots tags.
3. Rewrite the first paragraph as a direct answer.
4. Add a table, a list, and an FAQ.
5. Wait for reindexing after the update, then compare the report.

To make the technical side easier, try our robots.txt generator.

Still, patience is part of the plan. Citation data depends on refresh cycles, so wait a few weeks and use Compare before you judge a change.

How do IndexNow and robots.txt affect the report?

Microsoft names IndexNow as a way to tell search and AI systems quickly about content changes. When you update a page, the IndexNow protocol lets you announce the change right away.

That also helps most on sites that change often, such as price and stock pages. Fresh data can act as a freshness signal for citations.

Be careful with robots.txt as well. The report respects content controls, so a path you block will not appear in the data. For a broader view of bot management, read our guide to AI crawlers and robots.txt.

Local businesses should also keep their Bing Places for Business profile current, because local AI answers may draw on it, as Microsoft notes. You can also publish an llms.txt file as an extra guide for AI systems.

How do you build a monthly routine around the report?

However, opening the report once creates no value. Set up a fixed routine instead. We suggest a monthly review, because daily citation data swings and the monthly trend gives a healthier signal.

1. In the first week of the month, compare the last 30 days with the previous 30.
2. Note total citations and average cited pages.
3. List the five most cited and the five fastest falling pages.
4. Check the shift in Intents and Topics.
5. Pick three actions and measure the result next month.

Example calculation: if a site had 10 cited pages in month one and 14 in month two, its coverage is growing. Also, these numbers are only an illustration, not real client data.

Also track visits from AI separately with our guide to AI traffic in GA4. That way, you can also link citations to visits.

Why does the report matter for US and UK sites?

In the US and UK, Bing holds a smaller share than Google. Still, Copilot reaches desktop users directly through Windows and Edge. For B2B and office users in particular, this surface grows quietly.

So ignoring Bing data means missing a channel you have never measured. It is also free, with no setup cost.

How much data the report shows for your queries will vary from site to site. If the numbers look small, still do not panic. Early data from a preview product shows direction better than size.

In short, open the report, record a baseline, and compare it in three months. That single move is the cheapest measurement step you have in the AI era.

How does the report relate to ranking?

Classic search shows a position in a results list. But AI answers have no list. They have an answer, with sources placed inside or beneath it. So a citation is not a direct match for a ranking.

Microsoft makes this distinction clear. The metrics show citation frequency, not importance, ranking, or placement. A heavily cited page may not sit in the most visible spot of the answer.

So keep the two measurements apart. Use Search Console and Bing's search reports for organic ranking, and this report for AI visibility. Side by side, they show which pages are strong in both.

For instance, a page that ranks on page one but earns no citations may lack an answer-ready structure. The opposite can also happen. This comparison tells you where to invest.

Do freshness and indexing affect citations?

Microsoft's guidance lists freshness and accuracy among the traits of cited content. That suggests outdated pages may fall behind over time. Citation data can also shift with partner refresh cycles and model changes.

Also, a page cited today may drop tomorrow, and your content is not always the reason. User behavior, model updates, and freshness signals all play a role.

• Review pages with prices, dates, or regulations at set intervals.
• Show the update date on the page.
• Notify search engines with IndexNow after changes.
• Try an update before you delete an old page.

However, you do not need to refresh every page all the time. Evergreen guides need less care. Give priority to the pages that already earn citations.

How do service and e-commerce sites use the report?

On service sites, the report shows which questions you answer. A consulting site may earn citations with pages that explain the process and a price framework. Whichever page earns them, carry its structure over to your other service pages.

On e-commerce sites, category and guide pages tend to stand out. Structured data, clear spec tables, and current stock info matter on product pages. We cover this in GEO for e-commerce.

This table is a starting range based on field experience, not a guarantee.

How do you present this data to stakeholders?

First, executives do not want technical detail. They want three things: where we stand, why it matters, and what comes next. Present the report in that order.

1. Status: total citations and cited pages for the last 30 days, next to the previous period.
2. Meaning: which topics make you a source and which do not.
3. Action: three tasks for next month and the expected effect.

Second, do not oversell the numbers. State at the start that a citation is not a sale. That sets expectations on solid ground.

Add a footnote that the data is a sample and the product is in preview. That honesty also protects you when numbers swing later.

A monthly screenshot is enough for a simple record. Having past data in hand makes the trend easier to read.

How does the report fit a GEO strategy?

GEO is the work of making your content a likely source in AI answers. The Bing AI Performance report supplies the measurement leg. You finally see what your effort returns, at least on the Microsoft side.

The strategy has three steps: write answer-first content, keep technical access open, and measure the result. This report supports the third step. For the first two, our guide to SEO, GEO, and AEO differences offers a solid frame.

Also, do not swap GEO for SEO. A strong technical SEO base also helps in AI answers. Treat the report as a control panel that shows whether your base holds, not as a race table.

In short, measure, learn, and improve your content with what you find.

What mistakes do teams make with the Bing AI Performance report?

The most common mistake is equating citations with traffic. Next comes treating the data as complete when it is a sample. Last, some teams rewrite a content strategy after one phrase.

• Skipping site verification in Bing Webmaster Tools.
• Reading Citation Share like a competitive score.
• Assuming the report covers other AI platforms.
• Judging a change before waiting long enough.
• Never turning the data into action.

In other words, all of these come from turning measurement into the goal. The report cannot replace good content. It only gives that content a direction.

Another mistake is leaving the report to the SEO team alone. Content, development, and management teams all gain from it. If you want help, our SEO consulting team can combine Bing and Google data in one visibility view. To test the waters first, try our AI visibility checker.