GEO just got a scoreboard: Microsoft Clarity now tracks your AI citation share

What Citations Reporting actually tracks

Microsoft Clarity has been a free behavior analytics tool since 2020, used primarily for heatmaps and session recordings. The June 2026 update adds a distinct reporting surface for AI citation activity, separate from the human-visitor behavioral data Clarity was built around.

The four things Citations Reporting surfaces, per the 17 June 2026 announcement by Microsoft Advertising VP of Products Rukmini Iyer (about.ads.microsoft.com):

• Discovery events: when an AI engine retrieves your page as a candidate source for an answer
• Reference events: when your page is included in the context an AI model uses to construct a response
• Citation events: when your page is named as a source in the AI-generated answer a user sees
• Query attribution: the actual queries AI systems used to retrieve and evaluate your content before generating the answer

That last item is the most operationally useful. Seeing the query text tells you which questions AI engines associate your content with, which is rarely identical to the keyword clusters your SEO team has mapped. A fintech brand in Singapore may find its careers page being cited on "how to invest in Singapore in 2026" queries because that page contains a product list. Query attribution makes that visible; without it, you would never connect the citation event to the source page.

Citations Reporting is free for any site running Microsoft Clarity. Clarity itself is free. The only setup cost is the Clarity tag, which takes under five minutes to deploy and is already present on a substantial share of sites worldwide given Clarity's adoption trajectory since 2020.

Share of authority: the metric, the math, and the benchmark

The language Microsoft chose, "share of authority relative to other sources," is precise. It is a competitive metric, not an absolute one. Your share of authority on a query cluster is the fraction of AI-citation events where your content is the cited source, compared to total citation events across all sources on that cluster.

There is no industry-published benchmark for share of authority yet. The metric launched three weeks before this post was written. A reasonable working frame, drawn from share-of-voice research in paid search (where a 20-percent threshold often signals minimum-viable brand presence): treat anything below 15 percent on your core query clusters as a priority gap to close, 15-35 percent as competitive-but-improvable, and above 35 percent as a position worth defending through content freshness cadence.

These thresholds are directional. The more useful action is to track the number week-over-week and correlate it with changes you make to page structure, schema markup, and content update frequency. Share of authority is a lagging indicator of content-quality decisions made 4-8 weeks earlier, consistent with the indexing and re-grounding cadences reported by Web IQ teams.

For teams running analytics and insights alongside paid and organic channels, share of authority is the first GEO metric that fits cleanly into an existing measurement framework. It has a numerator, a denominator, and a competitive context. You can set targets for it the way you set targets for impression share in paid search.

Web IQ: the grounding layer behind your citation score

Citations Reporting shows you your score. Web IQ is what determines whether you have a score at all.

Microsoft Web IQ, introduced at BUILD 2026 earlier in June, is a suite of AI-native grounding application programming interfaces (APIs) that connect AI systems to fresh, authoritative real-world knowledge from across the web. It grounds Copilot and ChatGPT, among others. If your content is not accessible to Web IQ, it will not appear in the Clarity citations data for those two AI engines.

Web IQ operates differently from standard search indexing. Bingbot builds a keyword-relevance index from which it serves ranked links to human users. Web IQ fetches current, structured content that an AI model can incorporate directly into a generated answer. The quality requirements are also different: a page can rank top-3 in Bing search and still deliver poor grounding if its content is buried in JavaScript-rendered DOM, fragmented across multiple expandable sections, or lacks explicit factual claims with attribution.

Three content properties that Web IQ grounding rewards, consistent with Microsoft's own documentation and GEO research published by Princeton University (arXiv:2311.09735, November 2023):

1. Self-contained answer paragraphs: each paragraph reads as a complete answer without requiring the reader to consult surrounding context. AI models extract at paragraph level, not page level.
2. Named primary sources: statistics and claims attributed to a named source, not "research shows." AI models can evaluate and propagate source credibility; unnamed sources provide no credibility signal.
3. Explicit, crawlable markup: schema.org JSON-LD that matches the visible page content. Drift between schema and visible copy signals unreliable content to grounding systems.

If your AI visibility optimisation work has been focused on GEO-friendly content structure, Web IQ is the infrastructure that delivers the payoff. If it has not, Web IQ grounding is the first place to audit, ahead of content rewrites.

Brand Agents, Content Recommendations, and AI Max in Search

The 17 June 2026 Microsoft Advertising announcement packaged four tools together. Citations Reporting and Web IQ address measurement and indexability. Two more address conversion and campaign operations.

Clarity Content Recommendations is the actionable layer on top of Citations Reporting. It translates citation data into specific GEO suggestions: add a statistic to paragraph three, restructure section two as a self-contained answer block, add schema FAQ markup. This is the workflow integration that turns the metric into a task list.

Brand Agents are AI advisors deployable on your site through Clarity, with a WooCommerce integration available at launch. Microsoft's internal performance analysis (March 2026) found Brand Agent-assisted sessions show 2x higher conversion rates versus unassisted sessions. That figure is Microsoft's own benchmark across their Brand Agent customer base; treat it as a directional reference, not a guaranteed result for any individual deployment. Brand Agents are relevant to GEO because they close the loop: you earn a citation in an AI engine, the user visits your site, the Brand Agent continues the AI-assisted conversation that brought them there.

The Microsoft Advertising Model Context Protocol (MCP) Server is a read-only pilot for campaign management, usable inside M365 Copilot, Claude, and ChatGPT through Anthropic's Model Context Protocol (MCP). This is operationally separate from citation measurement but worth noting for teams running Microsoft Advertising alongside GEO programs: you can now query campaign performance, budget pacing, and keyword data through AI workflows without leaving the tools where you are already doing planning work.

AI Max in Search, the Microsoft Advertising AI-powered search expansion tool, will soon add search term landing page reporting, brand inclusions and exclusions, and term exclusions for asset optimization. These are measurement and control features for operators who have already turned on AI Max. Without landing page reporting, it is impossible to trace which expanded query variants are driving actual conversions versus assisted clicks. The addition of this reporting makes AI Max a more manageable tool for advertisers running it alongside independent measurement layers.

The 90-day action playbook for your GEO measurement data

Measurement without action is just cost. Here is how to convert Citations Reporting data into a structured program across a 90-day window.

Two market-specific notes. For Singapore and Malaysia clients: the IMDA AI Governance Framework (second edition) and Malaysia's MyDIGITAL blueprint both emphasize data provenance and content authenticity in AI systems. Brands that can demonstrate their content is accurate, schema-consistent, and freshness-maintained are better positioned for AI-grounding inclusion in regulatory and public-sector adjacent AI deployments, beyond commercial AI engines. For Australian and Canadian clients: regulatory frameworks around AI disclosure are moving fast (Australia's AI Assurance Framework, Canada's Directive on Automated Decision-Making). Brands in financial services and insurance in those markets should include AI citation accuracy in their content governance checklists, not just their SERP performance reviews.

This post is the measurement angle on GEO. The foundational how-to on GEO strategy, covering content structure, schema patterns, and AI-engine behavior, is at leapbuzz.com/blog/geo-generative-engine-optimization/. Read that first if you are starting from zero. Read this one when you have the setup in place and need a number to track.

If you want a structured audit of your current AI citation posture against these metrics, our AI visibility optimisation engagement starts with exactly that, and our SEO and AI visibility platform work covers the technical grounding layer. The analytics and insights practice connects the citation data to downstream conversion measurement so you are not optimizing a top-of-funnel number in isolation.