Microsoft has launched an AI Performance report inside Bing Webmaster Tools (public preview / beta) to help publishers understand how often their content is cited as a source in AI-generated answers across Microsoft Copilot, AI-generated summaries in Bing, and select partner integrations.

The feature is explicitly not a traffic report: Microsoft frames it as a first step toward “Generative Engine Optimization (GEO)” and repeatedly cautions that the new metrics reflect citation frequency, not rankings, prominence, or the role of any page in a particular answer.

The report introduces a small set of citation-centric metrics: Total Citations, Average Cited Pages, Grounding queries, Page-level citation activity, and a visibility trend line over time. The UI shows date presets (7D, 30D, 3M, plus Custom) and export controls (“Download” and “Download all”), but Microsoft also warns that parts of the dataset are sampled and may be refined retroactively as more data is processed.

Industry reaction has split into two themes reflected in the supplied sources: (1) the report is a meaningful “first-party” step because it exposes citations and internal retrieval phrasing (“grounding queries”); and (2) it is currently limited because it does not provide clicks/CTR/business outcomes (a point emphasized by Search Engine Land, Search Engine Roundtable, and social commentary).

What Microsoft Actually Shipped

Microsoft’s official announcement describes AI Performance as a “new set of insights” inside Bing Webmaster Tools that shows how “publisher content appears” across Microsoft Copilot, Bing AI summaries, and partner integrations, including visibility into which URLs are referenced and how citation activity changes over time.

The report is positioned as an early AI SEO feature: Microsoft notes that “visibility is not only about blue links,” and frames the launch as part of an emerging AI SEO tooling direction inside Bing Webmaster Tools.

From the interface screenshot published in coverage, AI Performance appears in the left navigation as a dedicated section labeled “AI Performance (BETA)”, and the top of the report states the citation sources as “Microsoft Copilots and Partners.”

Metrics, UI Elements, and Data Controls: Verified Inventory

Core metrics and Microsoft’s definitions

Microsoft’s official blog post defines the report as a consolidated dashboard and specifies what each metric measures:

Total Citations

Count of “citations that are displayed as sources” in AI-generated answers during the selected time frame; Microsoft emphasizes this highlights how often content is referenced without indicating placement or presentation.

Average Cited Pages

Average number of unique pages (unique URLs) from the site cited per day over the selected date range; Microsoft emphasizes it is aggregated across AI surfaces and does not indicate ranking, authority, or the role of a page in an individual answer.

Grounding queries

“Key phrases the AI used when retrieving content” referenced in AI answers; Microsoft explicitly states this is a sample of overall citation activity and will be refined as additional data is processed.

Page-level citation activity

Citation counts for specific URLs over the selected date range; Microsoft cautions that this reflects how often pages are cited, not importance, ranking, or placement.

Visibility trends over time

A timeline that shows citation activity changing over time across supported AI experiences.

Search Engine Land’s coverage restates the same metric set and emphasizes the same interpretive caveat: the metrics reflect citation frequency, not ranking/prominence or contribution to a specific answer.

UI elements and controls visible in the published screenshot

The screenshot included in coverage provides the most concrete evidence of report controls and layout:

• Date range presets: 7D, 30D, 3M, and Custom (a distinct “Custom” selector is visible).
• Metric cards for “Total Citations” and “Avg. Cited Pages,” each with a toggle/check control to plot on the chart.
• A dual-line time-series chart plotting “Total Citations” and “Cited Pages,” with a visible date axis and separate legend labels.
• A prominent inline notice: “The data shown below represents a sample of overall activity. Results may be refined as additional data is processed.”
• A “List By” section with two tabs: Grounding Queries and Pages, plus a query search box (“Search your query”) and table structure (column headers include “Grounding Query” and “Citations”).
• Export controls: a top-right “Download” button and a lower “Download all” control near the list/table region. The export format is not stated in the screenshot or Microsoft’s announcement.

Time windows, granularity, filters: what is specified vs. not specified

Specified:

• The report supports at least the visible date presets 7D, 30D, and 3M, plus Custom.
• “Average Cited Pages” is explicitly defined as a daily average over the selected range.
• The chart is a time series and appears (from the labeled x-axis dates) to be organized at day-level resolution, though Microsoft does not explicitly state the chart granularity in the announcement text.

Not specified in the accessible primary documentation:

• Maximum selectable date range, minimum/maximum “Custom” range, and any regional/device/search-vertical filters are not described in Microsoft’s announcement text and are not visible in the provided screenshot.
• Update cadence/processing lag is not stated in the Microsoft announcement; Microsoft only indicates that the dataset may be refined as more data is processed, and that grounding queries are sampled and under refinement.
• Sampling rate or minimum thresholds (for inclusion in the grounding query sample or page list) are not disclosed in the sources provided; multiple sources simply describe the dataset as a “sample” and/or “refined” over time.

Attribution Logic: How Citations and “Grounding Queries” Seem to Work

What Microsoft explicitly claims—and what it does not

Microsoft defines “citations” operationally as sources “displayed” in AI-generated answers, and it cautions that citation counts should not be interpreted as ranking, authority, prominence, or contribution.

Microsoft also states that Bing respects content owner preferences expressed through robots.txt and other supported control mechanisms, implying that existing crawling/preview controls influence whether content can be used and therefore cited.

However, Microsoft does not publicly specify, in the accessible first-party sources, whether:

• a citation count increments per answer impression, per answer render, per citation rendering event, or per unique answer instance,
• multiple citations to the same URL in a single answer are possible and how they are counted,
• citations are limited to web documents vs. other sources (e.g., licensed sources, knowledge bases),
• or how partner-surface attribution is differentiated (the UI aggregates “Copilots and Partners,” and social commentary notes there is “no way to distinguish” among surfaces).

Interpreting “Grounding queries” as internal retrieval phrases

Microsoft describes grounding queries as “key phrases the AI used when retrieving content,” and emphasizes the data shown is a sample and under refinement.

Search Engine Roundtable contextualizes the confusion by suggesting grounding queries are likely not the literal user query, but rather the phrases Bing uses to retrieve content (and speculates that Copilot may break longer user prompts into shorter retrieval queries). This is presented as interpretation/speculation, not a Microsoft-confirmed mechanism.

This aligns directionally with how Google describes AI Mode and AI Overviews behavior on Google Search Central: both AI features may use “query fan-out,” issuing multiple related searches across subtopics to build a response.

Social posts as practical interpretation signals

Ann Smarty highlights two practical implications: grounding queries are described as key phrases used to retrieve content (not prompts), and the report aggregates multiple AI surfaces without distinguishing them. She also reports Microsoft’s description of the core unit as “how often a specific URL … is visibly cited in an AI-generated answer.”

Harpreet Chatha’s post reframes grounding-query data as something to export and then reconcile with traditional search performance metrics (impressions/clicks/CTR) to estimate how many people might skip AI answers versus click classic results; this is proposed as a workflow rather than a confirmed reporting feature of AI Performance itself.

Contradictory reporting that must be treated cautiously

WebProNews states that Bing’s AI Performance report “tracks impressions, clicks, and engagement in AI responses” and that it “documents whether users ultimately click through to source websites.”

This conflicts with multiple other supplied sources that explicitly characterize the report as citation-focused and note the absence of click data:

• Search Engine Land says Bing Webmaster Tools “still won’t reveal” how citations translate into clicks/traffic and notes “without click data” publishers can’t tell if AI visibility delivers value.
• Search Engine Roundtable states “no click data.”
• The UI itself shows citation and cited-pages metrics but does not display clicks/CTR in the visible view.
• Given Microsoft’s own announcement enumerates citation metrics and does not mention AI clicks/CTR/impressions, WebProNews’s “clicks and engagement” characterization should be treated as unverified or at least not supported by the primary Microsoft announcement.

Why the Tool Matters: Use Cases, Strategic Actions, and Risks

Microsoft positions AI Performance as a transparency feature for publishers as AI becomes a more common way people discover information, shifting the optimization question from “rankings” toward whether content is referenced in AI-generated answers.

High-value use cases for SEOs and site owners

Because the report exposes which URLs are cited and which key phrases (grounding queries) are associated with retrieval, it supports several practical use cases consistent with Microsoft’s guidance and the social commentary:

Citation-based content auditing
Identify which pages are already being used as references and which pages/topics appear frequently; Microsoft explicitly suggests using cited pages and grounding query phrases to validate visibility and spot opportunities to improve structure and completeness.

Monitoring AI visibility trends over time
Use the visibility timeline to detect trends and changes over selected windows (7D/30D/3M/custom) and correlate them with content updates or technical changes.

Operationalizing “grounding query” intelligence
Export grounding queries (export affordances exist in the UI) and treat them as a retrieval-intent vocabulary: Harpreet suggests focusing on high-value brand + capability phrasing and using spreadsheet workflows to organize and interpret.

Freshness orchestration
Microsoft highlights IndexNow as a mechanism to keep content fresh across search and AI experiences by notifying participating search engines when content is added/updated/removed, framing freshness as important for inclusion and citation.

Local and entity accuracy
Microsoft notes that local businesses should keep business details current and points to Bing Places as an additional mechanism for maintaining accurate address/hours/contact information to support AI visibility for location-based queries.

Researcher-facing use cases

For researchers studying generative retrieval and citation behavior, AI Performance provides a first-party signal of which URLs are being cited and how citation activity changes over time—something the industry has frequently lacked. This framing appears in multiple supplied sources that describe it as the first time Bing Webmaster Tools shows citation frequency and page-level citation activity within AI answers.

Key risks and failure modes

Misattribution and over-interpretation
Microsoft repeatedly warns that citation counts don’t indicate ranking, prominence, authority, or the role of a page in any answer; treating “Total Citations” as a proxy for “AI share of voice” without that nuance risks incorrect strategic conclusions.

Sampling opacity
Both the interface notice and Microsoft’s definition of grounding queries describe sampling and refinement; that implies counts and query lists may shift over time and may not reflect the full universe of citation events.

Cross-surface ambiguity
The report aggregates citations across Copilot, Bing AI summaries, and “select partner integrations,” and social commentary notes there is no way to distinguish among these surfaces in the report. This makes channel-by-channel optimization or troubleshooting harder.

Overreliance without outcome measurement
Search Engine Land and Search Engine Roundtable emphasize the missing click/outcome layer, arguing citation frequency alone doesn’t reveal business value.

Privacy and data-collection considerations
Neither Microsoft’s announcement nor the accessible UI screenshot describes user-level data exposure; instead, the feature presents aggregated counts and sampled phrases. Meanwhile, Microsoft emphasizes compliance with robots.txt and other control mechanisms, implying that existing site-owner controls govern what content can be crawled and surfaced in AI answers.

Separately, Google’s official documentation underscores that AI features are part of Search and that site owners control crawling and preview via robots and snippet controls; this serves as a useful industry benchmark for how major engines frame AI controls.

Publishers’ broader concerns about generative summaries using content without consent are documented in ongoing regulatory and industry disputes (for example, European publisher complaints regarding AI Overviews).

Competitive Landscape and Comparisons

Bing’s approach vs. Google’s approach

Bing’s AI Performance report is a dedicated AI citation dashboard that answers “were you cited” and “for which pages/grounding phrases,” not “how many clicks did you get.”

Google’s official guidance takes a different stance: AI Overviews and AI Mode are treated as part of Search, and sites appearing in AI features are included in overall Search Console traffic (Performance report, “Web” search type). Google also documents how clicks/impressions/position are counted for AI Overviews and AI Mode.

In particular, Google documents that AI Overviews occupy a single position and that all links in the AI Overview are assigned that same position, which makes “citation prominence” difficult to infer from Search Console alone.

Third-party analytics: what they can and can’t replace

Web analytics tools can capture what Bing’s citation dashboard currently does not: on-site behavior, conversions, and downstream outcomes—but only once the user clicks through. Google Analytics describes acquisition reporting as a way to understand where website/app visitors come from, including referral traffic.

However, analytics tools generally cannot tell whether you were cited but not clicked; the Bing feature is explicitly designed to expose “reference without visit” patterns, which traditional referrer logs can’t see.

Key differences at a glance

Bing Webmaster Tools AI Performance

• Measures how often your content is cited inside AI answers
• Dedicated AI reporting dashboard (preview/beta)
• Page-level citation counts
• Sampled visibility into grounding/retrieval queries
• No click or CTR metrics currently shown
• Cannot yet segment by specific AI surface (Copilot vs others)
• No API access available yet

Google Search Console

• Measures impressions and clicks (AI interactions included in totals)
• No standalone AI dashboard
• No citation counter
• AI features merged into standard Search reporting
• Provides clicks, impressions, CTR, and average position
• Treats AI as part of overall Search ecosystem
• API available

Web analytics platforms (GA4 / Adobe Analytics)

• Measures behavior after a user clicks through
• No native AI citation tracking
• Tracks sessions, conversions, and referrers
• Flexible time segmentation and traffic analysis
• API access available
• Cannot see AI prompts, grounding data, or citation activity

Methodological Concerns and Recommended Best Practices

Methodological concerns implied by the sources
Sampling and revision risk
The UI warning and Microsoft’s grounding-query definition indicate the dataset is sampled and may be refined, meaning time-series comparisons should be treated as directional rather than exact counts of “all citations.”

Metric validity vs. business value
Multiple sources note the absence of click/outcome data, which means citation trends can’t be directly translated into revenue impact without combining other measurement sources.

Ambiguous unit of analysis
Microsoft defines citations as “displayed as sources,” and Ann Smarty highlights that Bing frames the URL count as “visibly cited,” but the sources do not describe whether counts are per answer-view, per user, per session, or per unique answer instance.

Recommended workflow for responsible interpretation

flowchart TD
  A[Collect AI Performance data\nTotal Citations / Avg Cited Pages\nGrounding Queries / Page-level citations] --> B[Export grounding queries & pages\nvia Download/Download all]
  B --> C[Cluster grounding queries\nby intent/topic/entity]
  C --> D[Map clusters to cited URLs\nand content types]
  D --> E[Content & technical actions\nstructure/clarity/freshness]
  E --> F[Notify updates (IndexNow)\n& ensure crawl controls aligned]
  F --> G[Monitor citation trend line\n(7D/30D/3M) for directional change]
  G --> H[Triangulate outcomes\nwith web analytics conversions\nand Search Console traffic]

This workflow reflects Microsoft’s suggested uses (review cited pages + grounding phrases, improve structure/clarity/freshness, use IndexNow) while explicitly adding the triangulation step that Microsoft and Google both recommend in different ways: combine Search Console-style measurement with analytics for on-site outcomes.

Practical best practices

Treat Grounding queries as “retrieval phrasing,” not as literal user prompts; Search Engine Roundtable highlights confusion here, and Google’s “query fan-out” description provides an analogous mental model for how generative systems expand a user’s ask into multiple searches.

Use page-level citation activity to prioritize improvements on pages that already demonstrate “citation eligibility,” but avoid assuming that higher citation counts mean higher authority or better ranking—Microsoft explicitly warns against those interpretations.

When you run experiments (content refreshes, structural edits, internal linking), rely on trend direction over time rather than single-day spikes, because the UI indicates that results can be refined retroactively and grounding data is sampled.

Pair citation signals with outcome tracking in analytics: GA4 acquisition reporting is designed to show where visitors come from, but it only observes visits after a click; this is exactly why citation visibility and analytics must be used together if you want both “influence” and “traffic.”

Recommendations for Microsoft to Improve AI Performance Reporting

Close the measurement gap with outcome signals

Multiple supplied sources argue the missing element is clicks/traffic value; Search Engine Land and Search Engine Roundtable explicitly emphasize “no click data,” and social commentary expresses uncertainty about what to do with citation numbers alone.

A concrete improvement would be to add AI-to-site clicks (and ideally CTR) in a way that preserves user privacy, similar to how Google documents click counting for AI Overviews and AI Mode.

Make sampling and data processing transparent

Since both the UI and Microsoft’s definitions indicate sampling/refinement, Microsoft should disclose (at minimum) whether sampling is applied to:

• only grounding queries,
• the entire dashboard below the chart (as the UI notice implies),
• or specific thresholds (e.g., low-volume queries suppressed).

At a minimum, add a sampling indicator per widget and an explanation of what can change over time, so analysts can avoid false “wins/losses.”

Add surface-level segmentation

Ann Smarty notes the inability to separate Copilot vs Bing AI summaries vs partner integrations, and the UI header aggregates “Copilots and Partners.” A filter for surface/source would materially improve actionability (e.g., diagnosing why a site appears in one surface but not another).

Provide API access and clearer export semantics

Search Engine Roundtable reports a Microsoft employee statement that the AI Performance data is “not yet available via the API” but is on the backlog.

Given the UI includes “Download” and “Download all,” Microsoft should document export formats and provide:

an API endpoint for daily citation series,
an API endpoint for page-level citations,
an API endpoint for grounding query samples (with sampling metadata).

Clarify attribution units and counting rules

Microsoft’s definition of citations as sources “displayed” is helpful but incomplete; the report should explicitly specify the unit (per answer render, per user, per session) and how repeated citations to the same URL are handled. This would address the “numbers seem unbelievable” concern raised in social commentary.

Integrate with governance and content controls

Microsoft already emphasizes that Bing respects robots.txt and supported controls; the report could surface diagnostics when a site has constrained eligible content (e.g., “X% of URLs blocked from AI surfaces by robots/controls”) so publishers can intentionally manage AI inclusion.

References

News Sources

• Search Engine Roundtable “Bing Webmaster Tools Rolls Out AI Performance Report (With New Design)” by Barry Schwartz Published on Feb 11th, 2026 at 7:51 am
• Search Engine Journal “Bing Webmaster Tools Adds AI Citation Performance Data” by Matthew Southern Published on February 9th, 2026
• Search Engine Land “Bing Webmaster Tools officially adds AI Performance report” by Danny Goodwin Published on February 10th, 2026 at 12:34 pm
• WebProNews “Microsoft’s Bing Webmaster Tools Unveils AI Performance Metrics as Search Giants Race to Quantify Generative Engine Traffic” by Eric Hastings Published on Wednesday, January 28th, 2026

Social Media Discussions

Ann Smarty:

Our very first AI visibility data source is here. 😱 Are you excited? Don't be 🙂

Bing Webmaster Tools just launched the very first AI visibility report we are seeing from a search / LLM platform, which, admittedly, is not very actionable (they promise to work more on that).… pic.twitter.com/CwNcFHTiTs

— Ann Smarty (@seosmarty) February 11, 2026

Harpreet:

Initial thoughts on the bing webmaster ai performance data are all around figuring out what to do with it.

1. Citations & Avg cited pages mirror regular organic traffic (B2B = spikes in the week, declines on the weekend).

2. Probably not a good thing if one or a few pages… pic.twitter.com/CkJRlZydtz

— Harpreet (@harpreetchatha_) February 11, 2026

Kevin Indig:

Your move, Google!

This is a screenshot of Bing Webmaster Tools' new AI performance report.

It also reports citations by grounding queries and by citations.

Obvs early days, but I love this as a start. pic.twitter.com/oUlHO9tB8F

— Kevin_Indig (@Kevin_Indig) February 10, 2026

Cyrus Shepard:

Best thing about the new Bing AI Performance report is seeing millions of brand citations before realizing that they led to almost zero traffic and/or conversions pic.twitter.com/UuAyVDB94g

— Cyrus SEO (@CyrusShepard) February 11, 2026

Ben Alfrey:

Guys im loving the AI Performance Report in Bing – We've just audited 37,335 total page citations & 22,807 total query citations across across 5 B2B SaaS Clients.

Here's what the data actually says about what gets cited:

1. Self-serving listicles are an AI citation hack. pages… pic.twitter.com/Y93pNyUhPb

— Ben Alfrey | B2B (@benalfrey) February 13, 2026

Editor’s Note: This article was created using a new AI system we are working on called “Autojourno”. Autojourno was given the above referenced articles from credible news sources and social media conversations and was tasked with creating the ultimate blog post / news item about this announcement. The article has been published under “Joebot” since it had minimal human oversight.

GET A FREE SEO REVIEW