Microsoft Clarity Citations & Web IQ: Measure AEO in 2026

Microsoft Clarity Citations is the first free analytics tool that shows which queries an AI generated before it cited your site, and it went generally available on May 13, 2026. For anyone doing answer engine optimization, it turns a black box into a dashboard: page-level citations, a competitive share-of-authority score, AI referral traffic, and the internal grounding queries that drive it all.

The timing matters. The same quarter, Microsoft announced Web IQ — a suite of grounding APIs that already feed Copilot and, per Microsoft, ChatGPT — and opened a read-only pilot of its Advertising MCP server. Taken together, these signal that the surface an AI cites is no longer a downstream side effect of your search rankings. It is its own channel, with its own metrics, and for the first time you can instrument it.

This guide covers what actually shipped in Clarity, the six metrics on the dashboard, the grounding-query gap that breaks traditional keyword auditing, why AI traffic converts so much harder than search, and the honest status of Web IQ and the Advertising MCP — what is live now versus what is merely announced. Every figure below is sourced and, where it is vendor-stated, labeled as such.

On May 13, 2026, Microsoft moved Citations in Clarity from preview to general availability, announced by product manager Ihab Rizk on the Microsoft Clarity blog. The feature is free with no traffic limits, and it needs no CDN or server-side integration. Setup is just the standard Clarity tracking code plus domain verification through Bing Webmaster Tools or Google Search Console.

The critical scope detail, easy to miss in the launch coverage: Citations reports the times your pages were referenced in AI-generated answers across Microsoft AI surfaces — Copilot and Bing generative search. It does not directly report citations from ChatGPT, Google Gemini, or Perplexity. The structural optimization patterns it reveals transfer across engines, but the dashboard itself is a window into Microsoft’s answer surfaces, not a universal AI-citation tracker. If you treat it as the latter, your reporting will quietly undercount.

The Citations dashboard tracks six metrics, per Microsoft Learn. The two that change how you think about AI visibility are Share of Authority and Grounding Queries — the rest are familiar in shape but scoped to AI answers rather than to search results.

The remaining three round out the picture. AI Referral Traffic counts sessions arriving from AI assistants — the bottom of the funnel, where a citation becomes a visit. Cited Pages breaks performance down to the URL level, so you can see which specific pages earn citations rather than just a domain total. And Trendlines tracks all of the above over time, which is what makes the dashboard useful for reporting rather than spot-checking.

Share of Authority is the one to watch in client reporting. Because it is calculated relative to every domain in the same grounding event, it gives you a competitive read — are you gaining or losing ground in the answers your buyers see — without needing any access to a competitor’s analytics. That is a meaningfully different instrument than a rank tracker, which only ever shows your own position in a list.

A grounding query is not what your visitor typed. It is the internal retrieval instruction the AI generates before it fetches web content. Microsoft’s own example: a user asks “What’s the best skincare routine for sensitive skin?” and Copilot internally queries something closer to “best skincare routine sensitive skin” or “dermatologist-recommended sensitive skin steps.” You were never optimizing for the second phrasing, because until now you could not see it.

This is where traditional SEO auditing breaks. In an analysis by SALT.agency, 80.5% of the grounding queries Clarity surfaces had no corresponding Ahrefs ranking data — only 2% were exact keyword matches, 17.5% showed meaningful keyword similarity, and 19% of pages were cited despite no traditional search signal at all. SALT.agency’s sample methodology is not fully disclosed and the figure was independently noted by Search Engine Journal, so treat it as directional rather than precise. Even discounted, the direction is the point: if you audit AI visibility using a keyword tool, most of the demand that actually triggers a citation is invisible to you.

There is a further layer the dashboard does not yet show. SALT.agency draws a distinction between grounding queries — the first-order retrieval instructions — and fan-out queries, the secondary retrievals an AI generates mid-reasoning to fill gaps. This is SALT.agency’s framing, not a Microsoft product term. Clarity currently surfaces only the grounding layer, which means even the best AI-visibility instrument available today is showing you the first move, not the whole game.

The practical takeaway is uncomfortable but clarifying: AI citation share and traditional page rank are now diverging signals. A page can rank nowhere and still be cited heavily because its passages answer the grounding query cleanly, or rank well and never be cited because its strongest material is buried below a weak intro. If your measurement stack only watches rankings, you are optimizing for a channel that is no longer the only one that matters. This is the gap our agentic SEO engagements are built to close — pairing answer engine optimization fundamentals with citation-share instrumentation rather than rank tracking alone.

Microsoft did not ship AI-visibility measurement in one launch. It shipped it in sequence across three dashboards, each measuring a different stage of how AI discovers and uses your content. Reading them as one funnel — rather than three disconnected features — is the mental model that makes the tooling actionable.

Stage one is the crawl. Clarity’s AI bot-traffic dashboard, launched January 21, 2026, shows which AI platforms — ChatGPT, Claude, Gemini, Copilot, Perplexity — crawl your site most often. This is an upstream signal, and an important caveat lives here: high bot activity does not guarantee citations. Being crawled is necessary, not sufficient. Stage two is grounding and citation, which is what Citations measures from May 2026. Stage three is the AI referral traffic that lands on your site when a citation converts into a click.

No other analytics platform maps all three stages end to end today, which is why this funnel is worth diagramming explicitly in any AEO reporting you build. The common failure mode is to celebrate stage one — “the AI bots are crawling us a lot” — and assume citations and traffic follow. They often do not. The bots index broadly; the citations are far more selective, governed by passage-level quality rather than crawl frequency. Measure each stage on its own, and the drop-offs between them tell you exactly where to invest.

The reason to instrument AI citations now, while AI is still under 1% of most sites’ traffic, is that the visitors it sends convert at a rate that dwarfs search. In a Microsoft Clarity study published November 6, 2025, across 1,277 domains tracked over one month, AI referral traffic converted at 1.66% for sign-ups versus 0.15% from organic search — roughly an 11× differential — and 1.34% for subscriptions versus 0.55% from search. This is the most recent Microsoft-published data on the question; cite it as November 2025 findings, not as a June 2026 result.

The platform breakdown sharpens the case. In the same study, Copilot-referred traffic converted at 17× the sign-up rate of direct traffic and 15× the search subscription rate; Perplexity posted the highest sign-up conversion rate of any AI platform tracked; and Gemini showed 4× direct and 3× search for subscriptions. Crucially, over 52% of the 1,277 sites already received convertible LLM referral traffic — so this is not an edge-case channel for a handful of publishers. It is a live business concern for most publisher-class sites, with growth to match: AI traffic in that dataset grew +155.6% over eight months, against +24.0% for search, +21.5% for social, and +14.9% for direct.

Read forward, the asymmetry is the strategy. A channel that is under 1% of volume but converts roughly 11× better is not a rounding error you defer — it is an option you buy cheaply now, before the citation real estate gets competitive. The teams that instrument citation share in 2026 will know which content earns AI visibility before the channel is large enough for everyone to fight over it. You can see how these AI referral traffic conversion rates play out across engines in our deeper analysis of AI referral traffic by platform.

On June 2, 2026, at Microsoft Build, Knut Risvik — Distinguished Engineer for Search & AI — announced Microsoft Web IQ, described as a suite of AI-native grounding APIs built for the agentic era. The architectural idea is a real departure: Web IQ returns passages and structured evidence objects rather than full web pages. The stated rationale is that models need information, not documents, and documents are often a poor proxy for it — returning evidence rather than pages cuts token load and latency.

Be precise about status. As of June 20, 2026, Web IQ is waitlist-only — accepting expressions of interest at its sign-up page, with no published pricing, no API documentation, and no general-availability date. Microsoft states Web IQ already powers grounding inside Copilot and ChatGPT, but that claim is vendor-stated and not independently confirmed. Treat the product as coming soon, never as available now.

The reason Web IQ matters for AEO even before you can touch it: if Bing now grounds answers at passage granularity rather than document granularity, then a weak opening paragraph can suppress citations from an otherwise authoritative article. That is a different optimization lever than “write better content” in the aggregate — it is about making the specific passage that answers a grounding query clean, self-contained, and near the top. Page-rank SEO optimizes the document; passage-level grounding optimizes the paragraph.

Largely overlooked in the Citations and Web IQ coverage, Microsoft Advertising expanded its MCP (Model Context Protocol) server to an open pilot on June 17, 2026. It brings live campaign data into the AI environments agencies already use — M365 Copilot, Claude, ChatGPT — so you can build custom analysis workflows grounded in real performance data rather than exported spreadsheets.

The boundary to state clearly: the current pilot is read-only. Write and campaign-management capabilities are not part of the open pilot as announced — you can query, not mutate. Named pilot partners include Mediaplus Performance, ClickTech, Kelkoo, Diginius, Optymzr, Groupon, Stagwell, Realtor.com, and Conversios. Capabilities and limits are as declared by Microsoft at launch.

The opportunity here is cross-channel. With read-only access to live campaign data inside the same AI environment you use to analyze organic visibility, you can correlate a topic’s Citation Share in Clarity against its paid-search investment — and start to see which organic citation topics complement your paid spend and which cannibalize it. That AEO-to-paid correlation is hard to do with disconnected dashboards, and it is exactly the kind of workflow that justifies the read-only pilot even before write access exists. It pairs naturally with the Bing Webmaster Tools citation share metrics Microsoft shipped the same month.

Most coverage treats Clarity Citations in isolation. In practice you are assembling a stack — each tool measures a different stage, covers different AI platforms, and offers different action capability. The table below maps the main instruments on one framework so you can see the coverage gaps. The “platforms” column is the one that exposes the biggest misconception: no single free tool reports citations across every AI engine.

Read the platforms column down and the gap is unmistakable: the bot dashboard sees every engine’s crawler, but the citation and referral tools are scoped to Microsoft’s own AI surfaces. There is no free, first-party instrument that reports your citation share inside ChatGPT, Gemini, or Perplexity directly today. The honest posture for 2026 is to instrument what Microsoft exposes — it is the most complete free stack available — and to treat the structural lessons (clean passages, strong intros, entity clarity) as transferable across engines, while measuring non-Microsoft citation share through separate means. Building that measurement layer is part of what our analytics engagements set up for clients, and it complements a full AI citation visibility audit methodology.