Bing Webmaster Tools AI Performance: How to Measure Copilot Citations in 2026

Bing Webmaster Tools now gives website owners something AI-search reporting has largely lacked: first-party visibility into how their pages are cited in Microsoft Copilot, AI-generated Bing summaries and selected partner AI experiences.

The AI Performance report tracks total citations, cited pages, grounding queries and citation trends. Bing has also expanded the report with preview features for Intents, Topics, Citation Share and Compare.

The important caveat is just as useful as the metrics themselves:

A Bing AI citation is not a ranking, click, conversion, authority score or exact user prompt.

The report is best used as a diagnostic layer between traditional SEO and business analytics. It can show which pages and themes are being used to ground AI answers, but you still need Search performance, analytics and conversion data to judge commercial impact.

This guide explains how to read the report correctly and turn it into a practical GEO measurement workflow.

What is AI Performance in Bing Webmaster Tools?

Microsoft launched AI Performance in Bing Webmaster Tools as a public preview on 10 February 2026.

According to Microsoft's launch announcement, the report shows how publisher content appears as a cited source across supported AI-generated experiences.

Bing's current AI Performance help documentation says supported surfaces currently include:

• Microsoft Copilot
• AI-generated summaries in Bing
• select partner AI integrations

This is not the same as ordinary Bing search reporting.

Traditional search reporting asks questions such as:

Which query showed my page?
How many impressions did it receive?
What was the average position?
How many clicks did it earn?

AI Performance asks a different set:

Was my content visibly cited in an AI answer?
Which page was cited?
Which grounding phrase was associated with that citation?
Which topics and intents are appearing?
How is citation activity changing over time?

That makes it useful for GEO, but only if the metrics are interpreted on their own terms.

What does a citation mean in Bing AI Performance?

A citation is a visible reference to content from your site in a supported AI-generated answer.

Bing defines Total Citations as the number of times your content was visibly referenced or shown as a source during the selected date range.

That tells you citation frequency. It does not tell you whether your source appeared first, whether a user clicked it, whether it was the most influential source, whether Microsoft considers your site authoritative, or whether the citation produced revenue.

Bing is explicit that AI Performance does not measure ranking, authority, importance or a page's role inside an individual answer.

So this interpretation is wrong:

We received 5,000 citations
therefore
we rank #1 in Copilot

A defensible interpretation is:

Our content was visibly referenced
5,000 times across the supported AI experiences
represented in Bing's reporting sample.

That distinction prevents GEO reporting from turning into another mystery score.

What are grounding queries?

Grounding queries are one of the most useful and most easily misunderstood parts of the dashboard.

Bing describes them as key phrases the AI used while retrieving content that was referenced in AI-generated answers.

They are not necessarily the exact words a person typed.

For example, a user might ask:

What should I check before moving an old PostgreSQL database
to a newer version with minimal downtime?

An AI system may retrieve information using grouped or reformulated phrases such as:

PostgreSQL major version upgrade
pg_upgrade downtime
PostgreSQL extension compatibility

Bing may surface a grouped phrase representing that grounding activity.

Therefore:

Grounding queries are retrieval/context signals, not a transcript of user prompts.

Bing's current help page says the phrases are grouped representations of citation activity and do not expose individual AI answers, exact prompts or why a specific page was selected.

Use grounding queries to identify themes and tasks your content is already helping AI systems answer, not as keywords to clone into new pages.

Why grounding-query-to-page mapping matters

Bing connects the Grounding Queries and Pages views so you can inspect:

Grounding query
        ↓
Which pages from my site were cited?

and the reverse:

Page
        ↓
Which grounding queries are associated with it?

One grounding query can map to several pages. One page can map to several grounding queries.

This is valuable because raw citation volume can hide two different situations.

Strong topical fit

Page:
postgresql-upgrade-guide

Grounding queries:
PostgreSQL upgrade
pg_upgrade
PostgreSQL EOL
database migration

That suggests a focused page covering several related retrieval needs.

Potential cannibalisation

Page A → mobile app cost UK
Page B → app development cost London
Page C → mobile MVP cost UK
Page D → mobile app pricing London

If the grounding-query mappings overlap heavily, the site may have a duplication problem rather than a content-volume opportunity.

Use the mapping as a content architecture signal, not merely a leaderboard.

What is Citation Share?

Bing's current AI Performance documentation includes Citation Share as a preview capability.

Citation Share represents the percentage of citations attributed to your site out of the citation space for a specific grounding query.

Conceptually:

All citations Bing reports
for grounding query X
        ↓
Your site's citations
        ↓
Citation Share

This adds context that total citations alone cannot provide.

Imagine:

The second query has more citations but a smaller relative share.

That does not make Citation Share a rank. Bing explicitly says it is not a ranking, traffic metric or quality score, and it does not expose competitor domains holding the remaining share.

A useful interpretation is:

Citation Share tells you how much of the observed citation activity for a grounding phrase is attributed to your site, not where your site ranks in an AI answer.

What do Intents add?

Bing is previewing intent classification for grounding queries.

Current examples in Bing's help documentation include:

• Informational
• Commercial
• Navigational
• Learn and Solve
• Research
• Local
• Comparison
• Planning
• Utility
• Creation
• Conversational

This matters because two sites with the same citation volume can have different commercial value.

For a software company, useful questions include:

• Are troubleshooting posts earning only Learn and Solve visibility?
• Are commercial comparisons appearing under Commercial or Comparison intent?
• Are local service pages receiving Local intent citations?
• Are service pages absent while only informational blog posts are cited?

Intent data turns citation visibility into a content-portfolio decision.

Bing notes that classifications are produced by AI/ML models and may be imperfect on ambiguous phrases, so use them as analytical aids rather than ground truth.

What do Topics add?

Topics group related grounding queries into broader themes.

For example:

Topic:
Payment Reliability

Grounding queries:
Stripe webhook retries
duplicate webhook events
payment idempotency
subscription reconciliation

Topics are useful because content strategy should generally operate at the cluster level, not one phrase at a time.

A site can then ask:

Which themes generate citations?
Which themes generate commercial citations?
Which important service themes have no AI visibility?
Which topic is overly dependent on one page?
Which topic contains too many overlapping pages?

That is closer to a useful GEO programme than chasing every retrieval phrase individually.

What does Compare do?

The Compare capability overlays a previous period against the current period.

That helps answer:

Did citation visibility change
after we updated this content cluster?

But it does not prove causation.

Bing warns that citation changes can reflect user demand, content changes, model or system updates, freshness, partner refresh cycles, and changes across the wider web.

So this is acceptable:

Citation Share increased in the period after the update.

This is not established merely by the dashboard:

The update caused the Citation Share increase.

Use Compare for observation and hypothesis generation.

Why can totals differ between views?

Bing says AI Performance is aggregated and summarised rather than a complete log of every citation event.

Its documentation warns that:

• not all citation activity may appear
• very sparse activity may not surface
• totals can differ between page, grounding-query and time-series views
• data represents a sample of overall activity
• results can be refined as additional data is processed

Use AI Performance for:

trend analysis
comparative analysis
content diagnostics

not financial-grade citation accounting.

Does a citation mean someone visited your site?

No.

Bing explicitly says citations do not represent clicks, traffic or engagement.

A useful measurement funnel is:

AI citation visibility
        ↓
Search / organic visibility
        ↓
Website visits
        ↓
Engagement
        ↓
Lead / purchase / business outcome

Bing AI Performance gives you the first layer. Analytics and CRM data provide the later layers.

If a page receives 20,000 Copilot citations but generates no qualified visits or assisted conversions, that visibility may still have brand value, but it should never be reported as 20,000 visits.

How is this different from Google Search Console's AI reporting?

Google Search Console and Bing Webmaster Tools now provide different windows into AI-search visibility.

A useful high-level comparison is:

Do not force the two systems into one artificial metric.

Build a combined reporting model:

Google AI visibility
+
Bing/Copilot citation visibility
+
organic traffic
+
analytics
+
conversions

This creates coverage without pretending the platforms measure the same thing.

A practical monthly Bing AI Performance workflow

Step 1: export the report

Bing currently supports CSV and Excel exports for AI Performance data.

Export:

• time-series citation data
• grounding queries
• page-level citations
• available intent/topic data
• citation share where available

Keep a dated monthly snapshot because preview features and available history can evolve.

Step 2: classify pages by business role

Do not analyse all URLs as one pile.

Assign each cited page a role such as:

Technical troubleshooting
Buyer guide
Commercial comparison
Service page
Local service page
Research/explainer
Case study
Documentation

Now citation data becomes commercially interpretable.

Step 3: calculate citation concentration

Ask how dependent AI visibility is on a small number of pages.

Track:

• top page share of citations
• top five pages' share
• citations by content cluster
• number of cited pages

High concentration can indicate a strong flagship asset, but it can also create fragility if one ageing article supplies most of the site's AI visibility.

Step 4: map grounding queries to pages

Look for three patterns.

Strong match

One good page appears for several closely related grounding phrases.

Action: maintain and deepen it.

Missing coverage

An important grounding theme exists but cited pages only answer it partially.

Action: improve an existing page or create one genuinely distinct supporting article.

Cannibalisation

Several near-identical pages appear for the same narrow intent.

Action: consolidate, differentiate or improve internal linking rather than publishing another synonym page.

Step 5: inspect intent mix

Create a distribution such as:

Informational      48%
Learn and Solve    24%
Commercial         12%
Comparison          8%
Local               5%
Other               3%

Then compare that distribution with business goals.

A software agency with zero Commercial, Comparison or Local visibility may have strong educational authority but a weak buyer-intent content layer.

Step 6: connect citation data to analytics

For the most-cited pages, review:

• organic sessions
• engagement
• service-page clicks
• CTA interactions
• lead submissions
• assisted conversions
• branded-search trends where available

Do not expect a one-to-one relationship. The purpose is to see whether AI-visible content participates in a useful customer journey.

Step 7: write a hypothesis, not a victory statement

Example:

Observation:
Our backend troubleshooting articles have high
Learn and Solve citation activity.

Hypothesis:
Bing sees the site as useful for practical diagnosis.

Action:
Strengthen internal links from those articles
to one relevant service page and publish a distinct
buyer-stage guide only where a real gap exists.

Success measure:
Commercial-intent citation presence grows
without reducing the technical cluster's visibility.

That is a measurable content strategy.

A compact GEO dashboard

A monthly dashboard can contain:

Avoid turning every available field into a dashboard mural. The report should help decide what to update, consolidate, expand or stop producing.

How should you improve a page that is already getting cited?

Bing's own guidance recommends clearer structure and stronger evidence rather than a special AI-only markup trick.

Microsoft recommends practices including:

• descriptive headings
• concise sections
• tables and FAQs where useful
• deeper subject coverage
• evidence and cited sources
• current, accurate information
• consistent entities across formats

The reasoning behind those recommendations is increasingly explicit.

Microsoft's May 2026 search engineering discussion describes a shift from retrieving whole documents only for ranking toward finding groundable information for AI answers: discrete, supportable facts with clear provenance.

That does not mean every paragraph should become a tiny SEO capsule.

It means important claims should be:

clear
specific
supported
self-contained
current

A strong article can still be nuanced and readable.

What if a page gets citations but no commercial result?

First determine the page's role.

A troubleshooting article can correctly earn high citations while producing few immediate sales. Its job can be:

Solve problem
→ demonstrate expertise
→ introduce relevant service path
→ create assisted trust

Improve the journey rather than stuffing the article with sales copy.

Useful changes can include:

• one contextual internal link
• a clear next buyer question
• a relevant comparison, cost or service page
• a CTA that follows naturally from the solution

The page should remain valuable even if the reader never becomes a lead.

What if an important page has no citations?

Do not assume a penalty.

Bing says sparse activity may not appear, and a page can be absent for many reasons.

Check:

1. Is the page indexable?
2. Is Bing discovering the current version?
3. Does it clearly answer a real task?
4. Is the information specific enough to ground an answer?
5. Is it current?
6. Are consequential claims supported?
7. Is it duplicative of another page?
8. Does it have a clear entity/topic focus?
9. Do internal links clarify its role?

Then decide whether the page needs improvement or whether the underlying demand simply produces little reported AI citation activity.

Indexing and freshness still matter

AI visibility does not replace search-engine fundamentals.

Bing says AI Performance only reflects content eligible for indexing and respects supported controls such as __INLINE_CODE_0__.

Microsoft also recommends current sitemaps and IndexNow to notify participating search engines when URLs are added, changed or removed.

For frequently updated technical and regulatory content, a useful pipeline is:

Publish/update article
        ↓
Update sitemap + accurate lastmod
        ↓
Notify IndexNow
        ↓
Bing recrawls/reindexes
        ↓
Updated content becomes eligible for retrieval
        ↓
Observe AI citation trends later

IndexNow does not guarantee a citation. It helps make content changes discoverable faster.

Five mistakes to avoid

Calling citations rankings

Bing explicitly says citation activity does not measure ranking or authority.

Calling grounding queries exact user prompts

They are grouped retrieval phrases, not prompt transcripts.

Treating Citation Share as market share

It is relative citation presence within Bing's reporting model, not product-market share or traffic share.

Claiming one content change caused a citation increase

Use Compare to observe change and investigate. The report does not establish causality.

Optimising only for citation count

A page can earn large informational citation volume and little commercial value. Connect visibility to buyer intent and business outcomes.

A content decision matrix

This is the bridge from measurement to publishing.

How should a research-first blog engine use this data?

AI Performance should not become a keyword vending machine.

A stronger loop is:

AI Performance
        ↓
Identify cited themes and gaps
        ↓
Compare against existing pages
        ↓
Research live SERPs and primary sources
        ↓
Decide whether to update, consolidate or create
        ↓
Publish only if a distinct need exists
        ↓
Measure citations + search + business outcomes

This prevents a common failure mode:

grounding query found
→ generate article
→ query variant found
→ generate another article
→ internal cannibalisation

The dashboard should make content production more selective, not more prolific.

FAQs

Does Bing Webmaster Tools show Microsoft Copilot citations?

Yes. Bing's AI Performance report includes citation activity across supported AI experiences including Microsoft Copilot, AI-generated Bing summaries and select partner integrations.

Are grounding queries the exact questions users asked Copilot?

No. Bing says grounding queries are grouped phrases associated with retrieval and citation activity. They do not expose exact user prompts or individual AI answers.

Does a citation mean someone clicked my website?

No. Bing explicitly says a citation is not a click, traffic event or engagement metric.

Is Citation Share an AI ranking?

No. Citation Share shows your site's relative share of reported citations for a grounding query. Bing says it does not represent ranking, traffic or a quality score.

Can Bing AI Performance prove an SEO update worked?

It can show that citation activity changed after an update, but it cannot establish that the update caused the change. User demand, model changes, freshness and changes elsewhere on the web can also affect citations.

Should I create an article for every grounding query?

No. Group related phrases by user intent and topic, compare them against existing content, and create a new page only when there is a genuinely distinct task or buyer need. Otherwise update or consolidate existing content.

Conclusion

Bing Webmaster Tools AI Performance makes GEO more measurable, but only if its metrics are used carefully.

The dashboard can tell you:

How often your site is cited
Which pages are cited
Which grounding phrases are associated
Which topics and intents appear
How your citation share changes

It cannot directly tell you:

Your AI rank
Exact user prompts
Clicks
Revenue
Why a model selected you
Whether one content change caused a trend

The most useful workflow combines:

Bing AI citation data
+
traditional SEO data
+
website analytics
+
business conversions

Then use the combined evidence to decide whether to update a page, strengthen a cluster, consolidate duplication or create a genuinely new article.

That is a sturdier GEO programme than chasing screenshots of citation counts.

If you need to connect technical SEO, AI-search visibility and content-cluster measurement into one workflow, Softotic's SEO and growth optimisation service can help, while web application development can support custom analytics and reporting integrations.

Sources and references

• Bing Webmaster Tools: AI Performance
• Microsoft Bing Webmaster Blog: Introducing AI Performance in Bing Webmaster Tools Public Preview
• Microsoft Bing Search Blog: Evolving role of the index — from ranking pages to supporting answers
• Microsoft Bing Webmaster Blog: Keeping Content Discoverable with Sitemaps in AI Powered Search