Microsoft today added branded query segmentation to the AI Citations dashboard inside Clarity, letting site owners separate the grounding queries that name a brand from the generic ones AI systems run when composing an answer. It is the third citation-side release from the same product in 25 days, and the first that addresses a question the previous two left open: whether measured AI visibility reflects existing demand or new discovery.

The change was published on the Microsoft Clarity blog under a Clarity Staff byline, dated August 3, 2026. It carries no version number and no pricing note, and it arrives as a filter rather than a new dashboard.

The reasoning Microsoft offers for the release is narrow. According to Microsoft, not all AI visibility comes from the same kind of query: some citations are driven by queries that reference a brand directly, while others come from broader, generic queries an AI system uses to help answer a prompt. Without a way to separate those signals, the company states, it can be difficult to understand whether performance reflects direct brand visibility, category discovery, or a mix of both.

That is a modest framing for a distinction that has been doing quiet damage to AI visibility reporting since the reporting began. A citation count aggregates two behaviours with almost nothing in common. One is an AI system retrieving a company's own pages because a user typed the company's name. The other is an AI system retrieving those same pages while answering a question that had nothing to do with the company at all. The first measures recognition already earned. The second measures reach into an audience that did not know the brand existed.

What the release contains

According to Microsoft, the Citations dashboard now includes a branded versus non-branded view across query analysis and filtering. The company lists four changes.

Branded labels in the queries card mark individual queries in the queries view, so brand-specific queries can be identified and what AI systems looked up becomes visible at a glance.

Share of Authority breakdown by query type splits the Share of Authority card by branded and non-branded queries, showing where authority is strongest: queries mentioning the brand versus more general ones.

Branded and non-branded filters apply across dashboard data, allowing comparison of visibility when an AI system looks up a brand directly against how it performs on broader topics.

More precise citation analysis is the stated outcome of the other three. By separating brand-led demand from generic discovery, according to Microsoft, brand strength can be assessed more accurately, discovery and consideration opportunities identified, and changes in citation performance interpreted with greater confidence.

The release is live in the AI Visibility dashboard as part of the Citations experience, according to the announcement. Microsoft did not describe the segmentation as a beta, which distinguishes it from the topic classification feature that preceded it.

What the sample data shows

The illustration published with the announcement is more informative than the bullet list. Five grounding queries appear, each tagged branded or non-branded, with a citation count and a share of authority figure attached.

The non-branded query "best all-mountain skis" records 2.5 thousand citations at 20.3 percent share of authority. Two branded queries follow: "alpine beginner ski boots" at 1.9 thousand citations and 24.2 percent, then "alpine ski insulated snow paints" at 1.6 thousand and 22.3 percent. Two further non-branded rows close the table: "top snowboard brands" at 1.1 thousand citations and 18.6 percent, and "waterproof ski jackets" at 986 citations and 15.9 percent. In each branded row, the token "alpine" is highlighted as the brand match.

Read as a set, the numbers demonstrate the pattern the feature exists to expose. Branded queries in the sample account for roughly 3.5 thousand of about 8.1 thousand total citations, or a little over 43 percent of volume. Yet the two branded rows carry the two highest share of authority figures in the table, while the highest-volume row, generic and unbranded, sits fourth of five on authority. A site reading only aggregate citation counts would see a healthy total. A site reading the split would see that its strongest authority position depends on users already knowing its name.

The inverse case carries more weight. A domain whose citations are overwhelmingly branded is not being discovered inside AI answers. It is being confirmed. Those citations follow demand that was generated somewhere else, by advertising, by press coverage, by word of mouth, and they will disappear the moment that upstream demand stops. Aggregate citation reporting has been unable to tell the two states apart.

Three releases, 25 days

The cadence is worth stating plainly. Topic Insights launched on July 9, 2026, giving every Clarity user competitive analysis around user-defined topics at no cost, at a point when rival AI visibility vendors reserved comparable functionality for paying tiers. Query Topics entered beta on July 22, 2026, grouping raw grounding queries into automatically generated themes scored on citation volume and share of authority. Branded query segmentation follows twelve days after that, and 25 days after the first.

Each of the three attacks the same underlying problem from a different direction. Citation data arrived in usable volume and immediately became unreadable. Topic Insights compressed it against a hypothesis a team supplied. Query Topics compressed it against patterns the data itself contained. Branded segmentation does not compress it at all: it splits the record along a line that changes what the totals mean.

The sequence sits on eighteen months of accumulated infrastructure. Clarity introduced AI channel groups on August 29, 2025, separating referral traffic from ChatGPT, Claude, Gemini, Copilot, and Perplexity. Clarity research published in December 2025 found AI-referred traffic had grown 155 percent across eight months while remaining below 1 percent of total visitors. Bot Analytics arrived in January 2026, exposing crawler-level request patterns, and a violations layer flagging robots.txt non-compliance followed in June.

Citation reporting itself began at Bing. The AI Performance dashboard entered public preview in Bing Webmaster Tools on February 10, 2026, and two days later Microsoft formally defined generative engine optimisation while positioning grounding as infrastructure beneath most major AI assistants. On March 23, 2026, grounding queries were mapped to the specific pages they cited. Clarity Citations reached general availability on May 13, 2026, bringing grounding queries, cited pages, share of authority, and AI referral trends to every project. The June 17, 2026 Web IQ announcement framed the whole stack as commercial infrastructure, with Microsoft stating that AI traffic is growing eight times faster than human traffic.

The distinction has a track record elsewhere

Branded versus non-branded is not a new axis in performance marketing. It is one of the oldest structural divisions in paid search, and the industry has spent the past year rediscovering how much it explains about AI.

Research published by Fractl and Search Engine Land on July 2, 2026 put a number on it. Across 1,010,848 high-volume keywords covering 379 brands in eight verticals, overall search demand fell 29 percent year over year, with fintech down 38 percent. The volume classified as non-branded, roughly 90 percent of the total, overlapped heavily with the informational content most exposed to AI displacement. Branded query share, by that analysis, was the asset that held its value.

Set against the Clarity release, the two findings interlock awkwardly. If branded demand is the durable asset in classical search, a brand whose AI citations are mostly branded may be reading its own resilience as growth. Conversely, a brand with heavy non-branded citation volume is capturing exactly the demand the Fractl data says is being redistributed rather than destroyed.

The methodological point has already been established on the traffic side. Similarweb research covered in June 2026measured downstream visits following ChatGPT recommendations across paired competitors in finance, travel, and beauty. Its qualifying condition excluded users who mentioned the brand name in their prompt, on the grounds that including them would contaminate the causal reading with prior awareness. Clarity's segmentation applies the same exclusion logic, except that it exposes both halves rather than discarding one.

Google has been instrumenting the same boundary from the advertising side. A native branded search toggle arrived inside AI Max on June 1, 2026, after expanded intent-based targeting was documented pulling branded and competitor queries into prospecting campaigns and muddying attribution. Attributed Branded Searches, which Google links to a $31 sales gain per query, went to global availability in late June 2026. Two large platforms are converging on the same conclusion from opposite ends of the funnel: brand-name traffic and category traffic are different products, and reporting that merges them produces decisions built on an average of two unrelated things.

What the split does not resolve

Several limits carry over unchanged from the previous releases, and Microsoft has not addressed them here.

The first is sampling. Microsoft acknowledged at the February 2026 Bing launch that citation query data represented a sample rather than a complete record. Segmentation inherits that property. A branded share calculated over a sample is a branded share of the sample, and the company has not published the sampling rate or its stability over time.

The second is classification methodology. Determining whether a grounding query is branded requires matching brand tokens against query text, and the announcement does not describe how that matching works. The sample table highlights the string "alpine" inside two queries, which suggests substring matching against a brand term. Whether the system handles misspellings, product line names distinct from the corporate brand, multi-word brands containing common nouns, or competitor brand names appearing in comparison queries is not stated. The edge case is well known in paid search: a query containing a car model name is arguably branded to the manufacturer and non-branded to a dealership. Nothing in the announcement indicates how Clarity resolves it.

The third limit is the one that has followed every release in this sequence. Share of authority and citation counts describe presence inside AI-generated answers. They do not describe what that presence earns. AI-referred traffic, by Microsoft's own December 2025 measurement, sat below 1 percent of total visitors in the analysed dataset. A more precise breakdown of a metric with no established conversion value is a more precise breakdown, not a business case.

There is also an asymmetry in what the feature can reveal. A site can now see the proportion of its citations that came from brand-name queries. It cannot see the proportion of relevant non-branded queries where it was absent entirely, because Clarity reports on the domain's own citation record. Topic Insights partially addresses that gap through competitor comparison, but the branded filter itself measures what happened, not what did not.

Why the classification line matters for marketers

For SEO and content teams, the operational effect is that AI visibility reporting can finally be diagnosed rather than merely tracked. A quarter-over-quarter rise in citations previously admitted several explanations, including a brand campaign, an algorithm change, seasonal category interest, or genuine content improvement. Splitting the trend by query type discriminates between at least two of them.

For brand teams, the branded share becomes something closer to a recognition index inside AI systems, measured at the retrieval layer rather than through surveys. For performance teams accustomed to running separate branded and generic search campaigns with different targets, the AI citation record now maps onto a structure that already exists in their reporting.

For publishers, the read is different again. A publisher's brand queries are its masthead, and heavy branded citation volume indicates that AI systems treat the outlet as a destination rather than a source. Heavy non-branded volume indicates the opposite: the outlet functions as raw material for answers about topics, with the attribution question that has dominated publisher disputes throughout 2026 attached to every one of those citations.

The classification also carries a practical hazard. Because brand strength inflates branded citation counts, and branded citations in the sample table carry higher share of authority, a dashboard reporting only totals rewards brands for demand they generated elsewhere. Removing that inflation is what makes the non-branded figure interpretable, and the non-branded figure is the one that measures whether content is doing work an AI system could not have done without it.

Microsoft has not published a roadmap for further segmentation, nor indicated whether branded and non-branded splits will reach Bing Webmaster Tools, where the citation tooling originated. The previous release, Query Topics, remains in beta with no general availability date.

Timeline

Summary

Who: Microsoft, through the Clarity analytics team, publishing on the Clarity blog under a Clarity Staff byline. The release affects brands, publishers, and content teams measuring how their pages are cited inside AI-generated answers.

What: Branded query segmentation across the AI Citations dashboard, comprising branded labels on individual queries in the queries card, a Share of Authority breakdown split by branded and non-branded query type, filters that apply the split across dashboard data, and the resulting separation of brand-led demand from generic discovery in citation analysis.

When: August 3, 2026, twelve days after the July 22 Query Topics beta and 25 days after the July 9 Topic Insights launch, with Clarity Citations having reached general availability on May 13, 2026.

Where: Inside the AI Visibility dashboard in Microsoft Clarity, as part of the Citations experience, applied to the grounding queries AI systems run when retrieving supporting information for a response.

Why: Aggregate citation totals merge two unrelated behaviours, an AI system retrieving content because a user named the brand, and an AI system retrieving it while answering a generic question. Sample data published with the release shows branded queries carrying the two highest share of authority figures in a five-row table, at 24.2 percent and 22.3 percent against 20.3, 18.6, and 15.9 percent for non-branded rows, which means a domain reading only totals cannot tell recognition it already had from discovery it newly earned.