Microsoft Clarity’s AI Visibility dashboard now labels which queries are branded and which are not, splitting a single blended Share of Authority score into two separate reads. Microsoft says the feature is live inside the dashboard’s Citations experience, in a blog post published August 3, 2026. The update turns one number that could move for two different reasons into two numbers, one for each.
Three pieces landed together, inside Citations:
- Branded labels in the queries card: individual queries are now marked branded, so brand-specific queries are identifiable at a glance.
- Share of Authority broken out by query type: the card separates branded results from non-branded results instead of reporting one blended figure.
- Branded and non-branded filters: dashboard data can be filtered by either, to compare visibility when an AI system looks up the brand directly against a broader, general topic.
Microsoft frames the point as separating brand-led demand from generic discovery, to assess brand strength, spot discovery opportunities, and read citation changes with more confidence.
What Is a Grounding Query?
A grounding query, in Microsoft’s own description, is a query an AI system uses to look up supporting information for a response. Behind an AI assistant’s answer sits a research step: the system runs its own lookups to find sources, and each lookup is a grounding query. Clarity’s Citations experience is where the new labels sit. They describe what the AI system went looking for, not what a person typed into a search box.
What Do Branded and Non-Branded Mean in Clarity’s Citations View?
In Clarity’s Citations experience, the two labels separate grounding queries where the AI system looks up the brand directly from queries about broader, more general topics. The queries card marks each grounding query with one of the labels, and the Share of Authority card reports a citation result for each group separately, instead of blending both into a single score. Microsoft does not publish the rule that decides which label a query receives, so the exact boundary between the two categories is not documented from outside the dashboard. The purpose is documented: Microsoft frames the split as separating brand-led demand from generic discovery, so a shift in citation performance can be traced to one side or the other.
One Score, Two Different Jobs
Before this update, AI Visibility reported one Share of Authority number for all citations, branded and non-branded together. That single figure moves when either half moves, so on its own it cannot say which one changed, or in which direction, when the total shifts. The breakout and the filters split that one blended read into two separate, answerable questions, in a tool that costs nothing to run.
We covered a related flattening problem in our coverage of two vendor studies on AI ghost citations. That piece looked at a different axis: a citation and a brand mention are separate outcomes that a single blended score also flattens into one number. The axis is different — citation versus mention there, branded versus non-branded here — but the failure mode is the same one.
A Free Dashboard, Not a Panel Study
Clarity’s split is read directly off what the AI system already does when it grounds an answer, with no recruited panel and no exposed-versus-control design. That is a different approach from a different kind of AI-search measurement we covered separately, which compares exposed and control panels to estimate lift. The two should not be read against each other: one is a free, always-on dashboard read, the other a panel-based lift study available to DISQO’s customers.
Trade coverage places the release in a sequence. PPC Land counts it as the third citations-side release from Clarity in 25 days, and separately notes that Query Topics entered beta on July 22, 2026. Query Topics is a separate feature, not part of this release.
What to Check in the Dashboard This Week
- Open the queries card and read which grounding queries carry the branded label, instead of guessing from query text alone.
- Read Share of Authority as two numbers, not one, and track branded and non-branded separately over time.
- Use the filters to see whether a shift in the old blended score came from the branded side or the non-branded side.
- Treat the branded label as a description of the AI system’s own lookups, not a count of how many people searched for the brand by name.