Microsoft Clarity, the free analytics tool many financial institutions already run beside Google Analytics, now has a dashboard that counts how often AI answers cite your pages. It reached general availability on May 13 and has had an update most months since: branded and non-branded query filters in August, custom brand terms on September 29, a topic report in beta. It costs nothing. A project administrator verifies the domain and the numbers start.
If you market a bank, a credit union or an asset manager, turn it on this week. Then read this before you report the numbers upward.
What it counts
Six things. Page citations, how many times pages from your domain were referenced in AI-generated answers. Share of authority, your domain's share of all citations against the other domains cited for the same queries. AI referral traffic, the share of sessions that arrived from an AI assistant. Queries, the retrieval phrases the AI systems used to find and evaluate your content before answering, with a filter for whether the phrase included your brand. Your cited pages. Trendlines for all of it.
Two of those are worth a marketing leader's time every month. The grounding queries show what the machines were looking for when they reached you, in their words. The branded filter splits that into prospects who already knew your name and prospects who asked about a category, which is the growth half. If the non-branded list is short, the AI systems are not reaching you when someone asks about auto loans in your market, or which fund suits a retirement account.
What it does not answer
Microsoft is careful about this, and the care is worth reading. The dashboard covers what Microsoft calls supported AI experiences. The company does not publish the list, and the data comes from its own grounding layer. ChatGPT, Gemini, Perplexity and Google's AI Overviews are not named anywhere in the announcements. Microsoft also says plainly that a citation count reflects how often a page was referenced and says nothing about its ranking or prominence in the answer.
So a rising citation count tells you an AI system fetched your rates page more often inside one family of assistants. It does not tell you what the answer said. A prospect asked which institution has the best used auto rate, or which fund belongs in a retirement account, and the answer named three institutions and a comparison site. Were you one of the three? Was the figure it quoted yours or a third party's stale copy of yours? Did that change after you fixed the rates page? None of the six metrics can say, because the unit of measure is your page, and the unit of growth is the prospect's question.
The question that measures growth
When a prospect asks an assistant the questions people ask before choosing where to bank, borrow or invest, how often are you named, how often are you the source, and how often does a third party answer for you. Asked the same way every month, across the assistants prospects use, so the number means the same thing in March as it did in October.
That is a different instrument from a citation dashboard. It starts from a fixed set of questions, runs each one through the major assistants, and records three things per answer: named or absent, source or third party, and which page or site the answer leaned on. A fix to the rates page shows up as a move in the named count the next month, or it does not, and either answer is useful.
How to use both
Turn Clarity on and read the non-branded grounding queries once a month. They are free evidence of what the machines wanted from you. Put the citation trendline in the appendix of the monthly report, with the coverage caveat written next to it so nobody upstairs reads a Microsoft-only count as the whole market.
Put the prospect-question read on page one. That is the number a CEO can act on, because it is the number a prospect sees.
What we do with it
We measure the prospect-question view with a fixed 48-question panel run monthly across the major assistants, and we read Clarity alongside it for every client that has it on. The free tool tells us what Microsoft's systems fetched. The panel tells us what the prospect was told. Both belong in the report, in that order of importance.
If you want to see what the assistants say when a prospect asks about your institution, book a 30-minute working session. Bring the five questions you most want an AI assistant to answer about you, and your Clarity project if it is on. You keep what we find.