Everyone seemed to have an opinion about Google's Generative AI Search optimization guide when it just dropped over the past weekend. That’s fair, it's Google, and anything they publish about how AI Overviews work is worth reading.
But while the SEO community was busy dissecting that document, something arguably more actionable quietly reached general availability: Microsoft Clarity's Citations feature.
Microsoft is the first platform to surface genuine first-party data from the grounding process: the retrieval stage AI systems run before generating any answer, and you can now access it in Microsoft Clarity directly in a specific AI visibility dashboard.
What the Microsoft Citation Dashboard tracks?
The Citations dashboard lives under Dashboards → AI Visibility → Citations in Clarity.
Compared to Bing Webmaster Tools' AI Performance report, I found Clarity noticeably more intuitive: you can drill down to page level, inspect the individual grounding queries behind each citation, and track how all of this moves over time through built-in trend lines.
Here's what each metric covers:
Page Citations
Page Citations show the total number of times pages from your domain were referenced in AI-generated answers during the selected time period, including multiple citations within the same answer.
It's an absolute number. But you can click on the trendline icon at the corner to also see the citation trendline. It is useful for tracking momentum, but best read alongside Share of Authority and Citation Rate rather than in isolation.
Share of Authority
This is my favourite metric in the dashboard. Share of Authority shows what percentage of all citations across queries where your domain appeared versus all other cited domains.
A high Share of Authority means AI is consistently picking you when your domain is in the mix.
Grounding Queries
The queries where your domain received at least one citation, plus the rate at which you were cited across all queries where you could have appeared.
The grounding queries your content is being retrieved for show what AI actually maps your pages to — and comparing that against your intended keyword targets will often surface gaps you wouldn't find through traditional rank tracking.
A page you optimised for one angle might be getting pulled into entirely different queries. Conversely, topics you thought you were covering may not be generating any grounding activity at all. Those gaps would be valuable insights when doing content optimization..
I also really like the built-in trend line view that lets you layer and compare performance across different queries over time. It’s useful for identifying which topics are gaining or losing grounding traction, and whether changes you've made to your content are actually shifting which queries your domain gets retrieved for.
My Cited Pages
This is a page-level breakdown of what's being cited, how often, and under which grounding queries. This is best for identifying what's working and what's being crawled but passed over on a specific page.
You can click into any URL in the table to filter the view down to that specific page and also see a trend line showing how its citation performance has moved over time. That makes it straightforward to tell whether a page is on a steady climb, plateauing, or losing grounding traction after a content change.
AI Referral Traffic
The share of your sessions that actually came from AI platforms, calculated as AI-referred sessions divided by total sessions. It closes the loop between "being cited" and "bringing in a visitor."
Where it becomes most useful is when you're trying to understand whether a content or structural change actually moved the needle: if citations are rising but AI Referral Traffic isn't following, the content is being used to build answers but not compelling users to click through.
One limitation worth keeping in mind: this metric is likely inferred from page referrer data, which means it won't capture everything. Clarity currently only identifies Claude, Microsoft CoPilot, Gemini, ChatGPT and Perplexity in their AIPlatform channel. Also, when the referrer is stripped, it will likely go as Direct.
So I would treat it as a directional signal rather than a precise count, where it’s useful for spotting trends, but probably an undercount of your actual AI-driven traffic.
Why Share of Authority is the metric I am most excited about
A lot of GEO tools on the market work by generating synthetic prompts, running them through AI systems, and scraping the outputs to look for your brand. There are two structural problems with this approach. First, you're inflating your apparent citation volume — those aren't real user queries, they're manufactured ones. Second, you're working with a confined prompt set and a handpicked competitor list, which means your Share of Voice numbers are inherently skewed.
If you pre-select five competitors to track, you'll only ever see yourself in relation to those five. Clarity shows you where you actually stand across real queries, including against competitors you didn't think to include.
For building an accurate baseline, the real grounding data is the only foundation worth trusting.
Unfortunately, at the moment you can only see your share of the total, but not a ranked breakdown of who the other domains are. It tells you how dominant you are in the mix, but not exactly who you're losing ground to. That's a meaningful limitation worth keeping in mind, especially if competitive intelligence is part of what you're trying to extract.
However, for building an accurate baseline, it is very helpful. For example, if your Share of Authority is flat but citation volume is rising, the whole space is growing but you're not outpacing it. If your Share of Authority climbs, you're genuinely gaining ground.
Connecting AI citations to real visitor behaviour
One angle I haven't seen discussed much: once you isolate AI Referral Traffic inside Clarity using the AIPlatform or PaidAIPlatform channel filters, you can apply Clarity's full behavioral toolkit to that segment.
That means heatmaps, session recordings, scroll depth specifically for users who arrived through an AI source. Users coming from AI answers tend to arrive with more context and clearer intent (they've already been informed by the AI's summary), so their on-page behavior often looks different from organic search visitors.
I'm interested in using this to identify friction points on pages that are receiving AI traffic but not converting — and to understand whether the content AI is citing is actually aligned with what those visitors want when they land.
It's also worth noting: Google Analytics recently added "AI assistant" as a specific channel, which signals that AI-referred traffic is increasingly being treated as a distinct segment worth tracking independently. Clarity's channel-level filtering is already ahead of that curve.
How I'd use Clarity’s AI visibility data?
Start by establishing a baseline, even if the numbers feel small.
If AI Referral Traffic is under 1% right now, that's still your starting point. GEO optimization compounds over months, not weeks — you need that early number to evaluate whether anything you do actually moves it.
Dig into Grounding Queries before touching your content strategy.
These are the queries AI ran before generating answers that cited you — and comparing them against your existing keyword targets will often reveal meaningful gaps between what you're optimising for and what AI is actually retrieving your content for.
It's also a useful reminder that GEO is still SEO at its core as Google in their generative AI search optimization guide. A solid content that ranks and gets discovered is still the foundation.
Grounding queries provide additional data points for you to understand what AI is looking for when it goes to retrieve an answer. It can help uncover topics (or subtopics) that inform your content strategy and structure.
Find your highest-cited pages and ask: what do they have in common?
Are they more directly structured? Do they lead with the answer rather than building up to it? Do they contain specific data points, named sources, or concrete examples that give AI something unambiguous to pull from? Do they cover a topic more completely than competing pages, or do they nail a narrow angle particularly well?
Those patterns are your template. Once you identify what's making those pages get selected, applying the same structural logic and content characteristics to lower-performing pages can be a quick win.
Prioritise by the gap between grounding activity and citation rate
The Grounding Queries tab surfaces Share of Authority (SoA) at the query level, which makes it possible to identify where you're under-performing relative to the broader competitive field. Queries where your Citation Rate is low but grounding activity is high are the ones worth prioritising first — AI is already retrieving content for those queries, just not yours.
At the page level, the same logic applies but with a more specific lens. If a particular page shows a strong Share of Authority across most of its grounding queries but one or two queries sit noticeably lower, those outliers are worth investigating. It likely means AI is finding the page for those queries but consistently preferring other sources, which could point to a coverage gap, a clarity issue, or a section of the page that isn't as well-developed as the rest. Those under-performing queries are where targeted content improvements will have the most direct impact on citation share.
Cross-reference bot activity with citation data
If a page shows high AI crawler activity but low citations, then it implies that AI is finding it and then passing on it. That's usually a clarity or authority problem. The content isn't structured or direct enough to be selected as a source. These pages are where I'd focus first.
Watch AI Referral traffic alongside citation counts
If citations are rising but referral traffic isn't, then it means getting visibility without bringing visitors through. The content is being used to build AI answers, but users aren't clicking to learn more — which might signal that the pages need stronger hooks or more depth.
It can also signal something more structural: informational content is being cited while your product or service pages aren't getting recommended. That's worth investigating further: whether it's a brand authority gap, a content coverage issue, or simply that the pages AI is selecting aren't the ones closest to a conversion decision.
Looking at which specific pages are driving citations versus which are driving referral traffic will usually point you toward the answer.
Real data in a speculative space
As generative AI search matures into its own distinct channel, something that Google Analytics is also now tracking separately, it is important for us to make data-informed decisions.
Microsoft Clarity brings in a new free feature with first-party data that is grounded in actual retrieval events rather than simulated queries or educated guesses.
For smaller teams and businesses that can't justify a hefty monthly subscription to dedicated GEO analytics platforms, this is a meaningful starting point. You can now establish a baseline and start tracking how citations, Share of Authority, and AI Referral Traffic move as you make content changes.