Microsoft Clarity AI Visibility: The Free Report That Shows Which Pages AI Cites

Microsoft Clarity AI Visibility is a free report inside Microsoft’s analytics tool that shows which of your pages get cited in AI-generated answers, how often, what the AI searched for before it picked them, and what share of those citations you hold against every other site cited alongside you. It reached general availability on May 13, 2026, and it costs nothing.

That last part is the reason to read on. Almost every conversation about AI visibility stalls at the same point: the client agrees the problem is real, asks what measuring it costs, hears a per-seat number, and defers it to next quarter. Microsoft took that excuse away. Most marketing teams already have Clarity installed for heatmaps and session recordings, and the citation report sits under a menu they have never opened.

What follows is what the report contains, what it leaves out, how to set it up, how it compares with the free reports from Google and Bing, and how to read a zero without drawing the wrong conclusion. It is a good tool. It is also easy to over-read.

What is Microsoft Clarity AI Visibility?

It is a set of dashboards inside Microsoft Clarity that report how AI systems interact with your site, and the one that matters for marketing is Citations: which of your URLs were used as sources in AI-generated answers, how many times, and for which underlying searches. Microsoft launched it to select customers on February 17, 2026, and made it generally available on May 13, 2026.

Clarity itself is Microsoft’s free behavioural analytics product, the one people install for heatmaps and session replays. AI Visibility is a newer branch of it. Microsoft’s Ihab Rizk described the February release as unifying “signals across Bing Webmaster Tools and Clarity” so a site owner can follow the chain “from bot crawling and indexing to how your content is cited in AI-generated answers and drives AI-referred traffic to your site.” Microsoft calls it an early step toward Generative Engine Optimization tooling, which is fair: a first-party window, not a finished product.

The commercial framing came two months later. Tim Frank, Microsoft’s corporate vice president for AI monetization, wrote on April 21, 2026 that the report exists because “traditional web metrics like rankings and clicks no longer tell the full story,” citing HUMAN Security’s 2026 State of AI Traffic report: automated traffic growing 8 times faster than human traffic, AI-driven sessions nearly tripling in 2025, agentic browser traffic up roughly 8,000% year over year. Those are HUMAN Security’s figures quoted by Microsoft, a vendor’s reading of the market. They are also why a free analytics tool now has a citation report in it.

What does the Clarity Citations dashboard actually measure?

Six things, per Microsoft’s May 2026 release notes: page citations, share of authority, AI referral traffic, the grounding queries behind each citation, a page-level list of which URLs were cited, and trendlines for all of it over time. It does not measure rankings, impressions or click-through rate. Microsoft says so directly.

The definitions matter, because two of them are easy to misread.

Page citations is the total number of times pages from your domain were referenced in AI-generated answers in the period, “including multiple citations within the same answer.” One answer that quotes you three times counts as three. And, in Microsoft’s words as reported by Search Engine Roundtable in May 2026, “Citation counts reflect how often a page was referenced, not its ranking or prominence within an AI-generated answer.” A citation buried eighth in a source list counts the same as the one the answer leads with.

Share of authority is the competitive number and the one most likely to end up on a slide. Microsoft defines it as the percentage of total citations attributed to your domain compared with other cited domains “within the same set of queries where your domain appeared.” Read that clause twice. It is not your share of AI answers in your category. It is your share of citations inside the queries where you were already cited at least once. Search Engine Journal’s Slobodan Manic, reading Microsoft’s documentation in August 2026, put it plainly: the metric “does not represent your share of the entire AI search market.” A site cited in two queries a month can post a strong share of authority and still be invisible for everything that matters.

Grounding queries are the most useful thing in the report and the least understood. They are not what the user typed. They are the searches the AI system ran on its own before it wrote the answer. Ask an assistant which tools show how visitors use a website and it may search several product names; Clarity records those as the queries behind your citation. It is the closest free view anyone has of how retrieval systems phrase your topic, which makes it keyword research in its own right.

AI referral traffic is separate on purpose: AI-referred sessions divided by total sessions. Citations happen before the click; referral traffic is the click. A page can be cited constantly and visited rarely, and Clarity shows both so you can see the gap.

Two additions arrived over the summer. On July 22, 2026, Microsoft added Query Topics, which groups grounding queries into automatically generated themes with citation count and share of authority per theme, exportable. On August 3, 2026, it split the dashboard into branded and non-branded queries, because, in Microsoft’s framing, “some citations are driven by queries that reference your brand directly, while others come from broader, generic queries.” The non-branded view is the honest one. It tells you whether AI finds your content when nobody asked for you by name.

Where does the data come from, and which AI assistants does it cover?

The data comes from Microsoft’s grounding layer, the retrieval system Bing operates for AI assistants, and it covers Microsoft’s own AI experiences plus partner platforms that use Bing to retrieve and ground answers. Microsoft does not publish a list of which partners those are. So the coverage is broad, and the boundary is not visible from the outside.

Grounding is the mechanism that connects a language model to live web content so its answer reflects the world rather than its training data. Jordi Ribas, Microsoft’s corporate vice president for search and AI, wrote on February 12, 2026 that “Microsoft grounding powers nearly every major AI assistant in the market.” Microsoft Advertising used a softer version in April: “most of the leading AI assistants.” Either way, this is why Clarity’s citation data is bigger than Bing. When an assistant’s retrieval runs through Microsoft, the citation lands in your report.

When it does not, it does not. Microsoft’s February 2026 methodology note calls the dashboard “a representative view of grounding and citation activity rather than a complete log of every reference.” Very low volumes may be excluded, totals can differ between the pages view and the queries view, and the stated design goal is “trend analysis and comparative insights over time, not precise accounting of individual AI answers or prompts.”

That is an unusually honest paragraph for a vendor to publish, and you should hold Microsoft to it. An assistant running its own retrieval stack for a query will not register. A number in Clarity is a real citation. The absence of one is a weaker signal than it looks.

How do you set it up and get to the first useful number?

Create a Clarity project, verify you own the domain by one of three routes, then open Dashboards, AI Visibility, Citations and filter to non-branded queries. Verification is by installing the Clarity tracking code, connecting Google Search Console, or connecting Bing Webmaster Tools, and it has to be done by a project administrator. Budget an afternoon, most of it waiting.

Three practical details from Microsoft’s own documentation will save you a support ticket.

1. Pick the right domain the first time. Microsoft’s May 2026 notes are explicit: for projects associated with more than one domain, “a single domain must be selected during setup for Citations reporting,” and “support for changing the selected domain after setup is not yet available.” If your marketing site and your app share a project, decide now which one you want measured.
2. You do not need the CDN connection. Clarity’s separate Bot Activity report wants an edge-provider integration. Citations does not. Domain verification is the only prerequisite, so a Search Console connection is enough to get citation data without touching the site.
3. Expect a lag. Data is refreshed daily “with a short processing delay,” and history is limited to the ranges shown in the dashboard. Do not judge the report on its first morning.

Once it is populated, the first number worth writing down is non-branded share of authority, and the second is the list of non-branded grounding queries behind it. Together they answer the only question that matters at this stage: when nobody typed your name, did the retrieval layer find you, and for what. Microsoft’s own customer example is instructive. A Digitas analyst on a health-sector client reported that “the biggest source of demand wasn’t from branded queries” but from exploratory searches around conditions and wellness, which sent the content team toward educational pages rather than brand pages. That is the report doing its job: moving effort, rather than counting.

Then set a monthly rhythm: compare branded and non-branded trends, find the pages and topics behind the biggest movement, review share of authority for each group, and choose one content or distribution action. If you have already built a prompt set the way we describe in our guide to AI prompt tracking, the grounding-query list is the first real-world check on whether that set matches how retrieval systems phrase your category.

How does Clarity compare with Search Console’s AI report and Bing Webmaster Tools?

They measure three different things and none replaces the others. Search Console reports impressions inside Google’s own AI features and nothing about queries or clicks. Bing Webmaster Tools reports citations and citation share in Copilot and Bing’s AI answers. Clarity reports citations across the wider set of assistants that ground through Microsoft, plus AI referral traffic on your site. Run all three.

The Google column is the one people expect to be strongest and is currently the thinnest. Google’s June 3, 2026 announcement lists impressions, pages, countries, devices and dates. It does not list queries, clicks or click-through rate, and Google said only that it is “adding additional metrics over time.” Search Console can tell you a URL appeared inside an AI Overview and how often; it cannot tell you for what, or whether anyone acted on it. Clarity cannot see Google’s AI surfaces at all, but for the assistants it does see it gives you the query. The two blind spots are mirror images, which is the argument for running both.

Bing Webmaster Tools and Clarity overlap, since both draw on Microsoft’s grounding data; Clarity is the wider footprint plus on-site referral data plus the branded split. If you only have time for one Microsoft report, it is Clarity.

Where none of the three go is the paid tier: prompt-level tracking, answer sentiment, whether the assistant described your product correctly, and how you fare against a named competitor on a defined prompt set across engines that do not ground through Microsoft. The free stack tells you that you were cited. It does not tell you what was said.

What does a zero in Clarity mean, and what does it not mean?

An empty “My cited pages” list means no page on your verified domain was recorded as a source in AI answers grounded through Microsoft during the selected period. It does not mean AI never cites you. It means the assistants Clarity can see did not, at a volume high enough to register, in that window. Treat it as a signal to investigate, not a verdict.

Work through the causes in order. First, verification: if the Citations dashboard itself is missing, the domain has not been verified by an administrator and nothing has been measured. Second, domain choice: if the project spans several domains and the wrong one was selected at setup, the report is faithfully measuring a site you did not mean to measure, and Microsoft does not yet let you change it. Third, date range: widen it, because low-volume activity can be excluded and the data lags. Fourth, the neighbouring metrics: if page citations, share of authority and grounding queries are all blank together, the simplest explanation is the real one.

Then ask what the zero is worth. A lot, if your category questions are ones people ask assistants, because the retrieval layer serving most of them is not finding you. Almost nothing on its own for a low-traffic niche site, where the “very low volumes may be excluded” clause can swallow the handful of real citations you did earn.

The edge case that catches people is the one-domain rule inside a multi-brand company. A group with four product sites in one Clarity project gets citation reporting for one of them. The other three read as zero forever, and someone eventually puts that zero in a board deck. If that is your structure, set up one project per domain before you look at a single number.

Is the free tool enough, or do you still need an audit?

The free tool is enough to establish whether AI retrieval finds you and for what. It is not enough to tell you what the answers say about you, how you compare on a defined set of buyer prompts, or what happens on the assistants that do not ground through Microsoft. Run Clarity first. Decide about an audit after you have read it for a month.

The sequence matters because the free report changes what a paid engagement should look at. A team that already knows its non-branded share of authority, its top grounding topics and its AI referral rate walks into an audit with the easy questions answered and the hard ones sharpened: why the category page is cited but the pricing page is not, whether the answer describes you accurately, and why a competitor holds the citation on the three prompts that drive pipeline. That is a cheaper audit and a more useful one.

Two things the report will not do, whatever Microsoft ships next. It will not tell you whether a citation is favourable; a page cited as an example of what not to buy counts the same as a recommendation. And it will not give you a stable single score, because citations vary answer to answer and day to day. We wrote about why a single AI visibility reading tells you nothing without its distribution, and the same applies to Clarity’s daily counts. Watch the trendline. Ignore the Tuesday.

Microsoft has said it will add competitor-gap views: where peers are cited instead of you, which topics they lead in, and recommendations to fill the gaps. When those arrive the free floor rises again. The gap between what a free tool reports and what an AI visibility audit concludes will not close, because the audit’s job is judgment: which citations matter to revenue, which answers are wrong, and what to change first. Clarity gives you the evidence. It still leaves you the decision.