How to Use Bing's AI Performance Dashboard for GEO
Bing Webmaster Tools AI Performance helps marketers see where their website is cited in supported AI answers. Use it to review cited pages, explore the grounding phrases associated with them, and observe citation trends.
This guide covers access, metric interpretation, the newer Intents, Topics, Citation Share and Compare views, and practical next steps. It also explains where Bing's platform-specific first-party data fits alongside Google reporting, cross-platform GEO analytics, GA4 and CRM data.
What is the Bing AI Performance dashboard?
AI Performance is a free report inside Bing Webmaster Tools. Microsoft documents coverage across Microsoft Copilot, AI-generated summaries in Bing and selected partner AI integrations in its launch announcement.
The report provides platform-specific first-party data about visible citations. It is not a complete or authoritative view of ChatGPT visibility or citations, and should not be described as ChatGPT analytics.
A citation shows that a page was visibly referenced. It does not show every page discovered or retrieved during answer generation, and it does not establish that someone visited your website.
Why use it?
It gives you a practical starting point for reviewing pages already appearing as sources in supported experiences. You can use those observations to choose what deserves closer attention, while keeping citation activity separate from search traffic and commercial results.
How to access the AI Performance dashboard
Start with a verified site in Bing Webmaster Tools. If you already have access, select the relevant property before opening the report.
Step 1: Add and verify your site
Open Bing Webmaster Tools and sign in. Import a verified property from Google Search Console, or add the site manually and follow the ownership-verification instructions.
Microsoft's verification guide covers XML file upload, a homepage meta tag, DNS verification and DNS auto-verification through supported providers. Choose the method your website team can maintain.
Step 2: Open AI Performance
Select AI Performance in the navigation. Review the citation timeline and the Grounding Queries and Pages views. Use the newer analysis views where available; Microsoft describes Intents, Topics, Citation Share and Compare as preview capabilities.
Step 3: Choose a reporting period
The documented ranges are 7 days, 30 days, 3 months and a custom range within available history. Choose a period with enough activity to review, and check the latest available date before making comparisons.
What each metric means and how to use it
Read citation volume, daily page breadth and relative citation presence separately. They answer different questions.
Total Citations
Total Citations counts visible source references during the selected period. It is a citation count, not a count of people, clicks or sessions.
Use it to follow reported volume over comparable periods. A rise is worth investigating, but could reflect changing demand or system behavior as well as changes to your pages.
Average Cited Pages
Average Cited Pages is the average daily number of unique URLs cited from your site. For example, 4, 6 and 5 unique cited pages across three days would average 5.
Use it to observe whether citations involve a broader or narrower set of pages. A low value does not diagnose weak structure or authority. Check the Pages view before drawing conclusions about which URLs receive citations.
Grounding Queries
Grounding queries are grouped retrieval phrases in sampled citation data, not exact user prompts. Use them as topic clues, not a transcript of what people asked.
An unexpected phrase can suggest a page to review. Open the associated pages and judge whether their information is relevant and accurate. The report alone cannot show whether a specific answer misunderstood your content.
Page-level citation activity
The Pages view shows citation counts by URL. Select a page to inspect its associated grounding queries, or select a grounding query to inspect the cited pages.
Compare the pages relevant to your marketing priorities. If a guide receives more citations than a product page, record that difference without treating it as proof that AI systems inherently prefer guides. The pages may address different needs and demand.
For a useful page review, ask: Is the information current? Is the page clear about the product or subject? Are important reader questions answered? Is the evidence behind its claims easy to check?
How to use Intents, Topics, Citation Share and Compare
Microsoft's June 2026 update adds context and comparison to the original report. Use these views to narrow your review rather than to assign a universal score.
Intents
Intents classifies grounding queries by broad query intent. The labels are machine-assigned and can be imperfect.
Review whether the categories shown match the audiences you want to help. A commercial label does not establish buying intent, a lead or revenue.
Topics
Topics groups related grounding queries into broader themes. Like Intents, it uses evolving classification systems.
Use themes to select relevant areas for closer review. Check the underlying phrases before acting on a broad label. For topic-based page analysis, Microsoft's help directs users to export the data and filter it.
Citation Share
Citation Share is your site's percentage of citations across all sites for a particular grounding query. It differs from total volume: a high count can coexist with a low share.
Review share alongside volume for the same query. It is not traffic share, a ranking or a quality score, and it does not identify the competing domains.
Compare
Compare overlays a previous or custom period on the current chart. Start with equal-length periods, such as the latest complete 30 days and the previous 30 days.
Use it to observe changes after a page update or across reporting periods. It does not isolate the effect of an edit; demand, model changes and other sources can also change.
How to filter and export the data
Filter by a query or page, then export the active view. Query and page filters cannot be applied together. CSV and Excel exports support query, page and time-series analysis.
Keep the property, date range and active filter with each export so you can compare like with like later. Do not add query and page totals together as if they were separate pools of citations.
How Bing AI Performance compares with cross-platform GEO analytics
Bing reports citation activity from its supported experiences. Cross-platform tools let you review a defined set of prompts across multiple engines and compare brands and sources. Choose them for that different scope, rather than assuming either dataset is a complete record of AI Search.
What Bing AI Performance cannot tell you
Microsoft's AI Performance help explains that the report is aggregated and sampled. Sparse activity may be absent, counts can differ across views, and processed data may be refined.
Treat an empty view as a reason to check the period and available data, not evidence that a page has never been cited. Likewise, the dashboard cannot explain why one source was selected over another.
The report does not provide individual conversation transcripts, complete ChatGPT reporting, on-site behavior or commercial attribution. Citation Share adds relative context without exposing competitor domains.
Data refreshes daily with a processing delay; use the history offered in your dashboard. Avoid assuming a fixed delay or a universal retention window.
How to turn Bing AI Performance data into useful next steps
Use the report to decide what deserves review. The goal is to make useful, accurate content easier for readers to understand, then observe the reporting afterward.
1. Choose a relevant topic or page
Start with a subject that matters to your audience. Use a grounding phrase, Topic or cited URL to select a small set of pages. Review those pages directly rather than reconstructing an exact question from a grouped phrase.
2. Check what the reader still needs
Look for outdated information, unclear explanations, missing examples or claims that need evidence. Compare related pages for useful coverage, not a supposedly winning format. Add a table or FAQ only when it makes the information easier to use.
If a priority page has little reported activity, check its accessibility and indexing with your website team before assuming a content problem. Being indexed does not guarantee citation.
3. Make a focused improvement
Fix the specific reader problem you found. Record the URL, change and date. Expanding a page is useful when it answers a real need; adding length, schema or a particular page structure is not a promise of more citations.
4. Observe what changes
Review comparable periods after enough new data is available. Note citation volume, pages and relevant query-level share, then check whether the pattern persists. Report movement as an observation, not proof that the edit caused it.
Validate any pattern separately on other AI engines. Bing's results do not establish a universal source-selection rule.
How Bing AI Performance fits into your AI Search measurement stack
Use each source for the question it can answer. Google and Bing provide valuable platform-specific first-party data; cross-platform analytics provides a unified view across tracked engines; website and customer data add downstream signals.
Google Search Console: Use Google's generative AI reporting for impressions and page visibility in Google Search surfaces such as AI Overviews and AI Mode. Google's announcement describes an impression-focused report, not separate AI click, CTR or query reporting.
Bing AI Performance: Use it for citation and grounding context across Microsoft Copilot, Bing AI summaries and supported partner integrations.
Cross-platform GEO / AI Search analytics: Use it to compare visibility, competitors, citations, sources, sentiment, brands, markets and trends across tracked engines. Keep the tracked prompt set and coverage in mind when interpreting results.
GA4: Review identifiable referral sessions and on-site behavior. A missing identifiable referral does not prove that AI played no part in discovery.
CRM, forms and sales calls: Add self-reported discovery, pipeline and closed-won outcomes. Keep those signals distinct from referral attribution and citation counts.
For enterprise marketing teams and agencies, Superlines provides AI Search analytics across brands, markets and engines, with competitive and source analysis and Action Center recommendations. It complements Bing Webmaster Tools and Google Search Console; it informs and measures content work rather than replacing an editor or CMS.
For a practical report, show these signals side by side with their labels and limits. Do not turn a citation count into a revenue estimate or expect the datasets to reconcile one to one.
Common questions about Bing AI Performance
Is Bing AI Performance free?
Yes. You can access the report through Bing Webmaster Tools with a verified site.
Is it ChatGPT analytics?
No. Microsoft documents Copilot, Bing AI summaries and selected partner integrations. The report should not be treated as a complete or authoritative view of ChatGPT visibility or citations. Use a dedicated AI visibility tool to assess ChatGPT across a defined set of tracked prompts.
How far back does the data go?
Use the historical ranges available in your property. Do not assume the February 2026 launch date defines the earliest available data or that every property has the same retained history.
Is the data real-time?
No. Microsoft documents daily refreshes with a processing delay, without specifying a fixed 24–48-hour delay in the current AI Performance help.
Can I see the exact prompt or Copilot conversation?
No. The report gives aggregate citation data and grouped retrieval phrases, not exact user prompts or individual conversations.
Why do citation totals differ between views?
Sampling and aggregation mean query, page and timeline counts need not match. Keep filters and periods consistent, and use exports for closer comparison without treating them as a complete event log.
Can Citation Share identify my competitors?
No. It shows your relative citation presence for a grounding query, without naming the domains holding the remaining share.
Your first dashboard review
Open a complete reporting period, select one topic relevant to your audience, and inspect its associated pages. Export the view and choose one concrete reader problem to fix. Review subsequent citation patterns alongside cross-platform visibility, identifiable visits and customer outcomes.