Microsoft published five example query patterns for the Clarity MCP Server on August 18, 2026, aiming to move users of its free web analytics tool beyond single-metric dashboards toward compound, AI-assisted questions that combine behavioral signals, segments and time comparisons in one prompt.

Credited to Clarity Staff on the Microsoft Clarity Blog, the post does not introduce new product functionality. Instead, it documents five prompt structures that marketers and analysts can send through an AI assistant connected to the existing MCP Server, along with the reasoning behind each query and a short set of authoring guidelines for building further prompts. Its publication date places it more than fourteen months after Clarity's MCP Server itself first launched, on June 4, 2025, which PPC Land reported at the time as an early example of an analytics platform exposing its data through the Model Context Protocol rather than a conventional dashboard export.

What the announcement covers

Clarity Staff frames the post around a problem familiar to anyone who has built a marketing dashboard: the most useful analytics questions tend to be narrow, conditional and multidimensional, and answering them through traditional tooling means stitching together several exports, filters and reports before a pattern becomes visible. An open standard that Anthropic introduced in November 2024 to let AI systems query external data sources through natural language, the Model Context Protocol changes that workflow by letting a single conversational prompt pull structured behavioral data directly into an AI tool.

The five patterns the post walks through are as follows.

Pages with high scroll depth but low conversions

Surfacing pages where users scroll deeply into content without completing a conversion action is the first pattern's aim, and the post's example prompt puts it directly: "Show me pages with average scroll depth above 75% but conversion rate below 5% for the last 30 days." According to Clarity Staff, a page fitting that profile likely has compelling content that keeps visitors engaged, but a call-to-action that is either poorly positioned or insufficiently persuasive. Repositioning buttons, simplifying forms and testing more direct copy are the responses the post suggests, though it does not claim any of those interventions as guaranteed fixes.

Sessions with multiple negative engagement signals

Rather than examining any one signal in isolation, the second pattern combines three behavioral indicators, rage clicks, excessive scrolling and quick backs, into a single filter. Clarity Staff argues that single metrics only tell part of the story, and that sessions exhibiting all three signals within a short window are more reliably linked to genuine user frustration than sessions showing only one. "List sessions where users exhibited rage clicks AND excessive scrolling AND quick backs in the last 7 days," the post's example prompt states. According to the post, this compound filter is particularly useful on checkout or signup pages, where unclear instructions or excessive form fields tend to produce this cluster of signals together.

Segmenting engagement across multiple dimensions

Breaking engagement metrics down simultaneously across device type, geographic region and browser is the aim of the third pattern, an approach the post argues reveals differences that aggregate averages hide. The post frames the query this way: "Show average engagement time for mobile users from organic search in North America vs Europe, broken down by browser." Clarity Staff frames this as a way to detect, for instance, that a responsive design performs unevenly across regions or browser engines, information that a single blended engagement figure would obscure entirely.

Shifting from a snapshot view to a comparison across time, the fourth pattern asks for pages where average session duration changed by more than a fixed threshold between one period and the previous one. "List pages where average session duration changed by more than 20% over the past month compared to the previous month," reads the post's example prompt. According to the post, rising trends can indicate a successful content or design change, while falling trends flag areas that need renewed attention, information that static, single-period metrics cannot surface on their own.

Combining session recordings with metrics for deeper insight

Connecting quantitative filtering to Clarity's session recording feature is what the fifth and final pattern does, asking the assistant to fetch recordings specifically for sessions matching a metrics-based condition, in this case low scroll depth paired with a high exit rate. The post's example prompt: "Fetch session recordings for pages with low scroll depth and high exit rate last week." Clarity Staff writes that metrics alone do not always explain why users behave a certain way, and that linking a recording directly to the metric that triggered its selection closes that explanatory gap.

Technical constraints that still apply

None of the five patterns changes the underlying technical limits that have governed the Clarity MCP Server since its June 2025 launch. Those limits, which PPC Land documented in its original coverage of the server, remain unchanged as of this publication: Microsoft's Data Export API permits ten requests per day per project, each request is limited to three days of historical data, and a query can span a maximum of three dimensions at once. Those three constraints matter directly for how the newly published patterns can be used in practice. Comparing a full month of session duration against the prior month, the fourth pattern requires stitching together data that exceeds the three-day window covered by a single request, meaning any user attempting that exact prompt will need to issue and combine multiple calls rather than retrieve the comparison in one. Similarly, the third pattern, which segments by device, region and browser simultaneously, sits at the three-dimension ceiling; adding a fourth axis, such as traffic source, would exceed what a single query can return.

Installation requirements for the MCP Server have also not changed. Running it requires Node.js version 16 or higher, an active Clarity project and an associated Data Export API token, which Microsoft's documentation, as PPC Land previously reported, instructs users to generate through the Data Export section of their project settings and store securely. Both temporary execution through npx commands and permanent installation are supported, and configuration can happen within Claude for Desktop, Cursor or other MCP-compatible clients through standard JSON files.

Illustrative data, not measured results

Each of the five sections in Microsoft's post includes a table labeled "Mock output," showing example figures such as scroll depth percentages, rage click counts and session duration trend percentages. These tables are explicitly presented as illustrative placeholders rather than results drawn from actual Clarity customer data or any named website. Nowhere does the post attribute the sample figures to a specific study, dataset, time period or set of properties, and none of the numbers in those tables constitutes a measured finding about real user behavior. What the tables demonstrate is the shape of the output an assistant would return for each query pattern, not a factual claim about how any particular page or segment actually performed.

That distinction sets this post apart from other analytics research Microsoft has published through Clarity. In December 2025, for example, Clarity's research team released findings, covered separately by PPC Land, showing that AI-referred visitors across a dataset of more than 1,200 publisher and news websites converted to sign-ups at 1.66 percent, compared with 0.15 percent for organic search, a gap the company attributed to Ravi Yada, Director of Product at Microsoft Clarity, and grounded in a named, measured dataset. The August 18 post makes no equivalent claim; its purpose is to show query structure, not to report a finding.

Guidance for building additional queries

Beyond the five worked examples, Clarity Staff closes the post with seven general principles for constructing further MCP queries. These include combining multiple metrics rather than relying on any single indicator to surface patterns, segmenting across dimensions such as device, browser, location or traffic source to reveal differences that aggregates hide, incorporating time-series comparisons to detect trends rather than relying on static snapshots, filtering sessions by behavioral triggers to focus on high-priority areas, using clear and specific language when writing prompts to improve output accuracy, documenting and reusing queries to maintain consistency across a team, and visualizing outputs to make results easier to interpret quickly. None of these principles is new to analytics practice broadly; what the post frames as new is applying them through a conversational, prompt-based interface rather than a traditional report-builder.

Context within Microsoft's broader Clarity roadmap

Arriving against a backdrop of sustained feature activity inside Clarity over the preceding fourteen months, the MCP query patterns post lands in territory where most of that activity has concentrated on AI-related measurement rather than the core MCP Server itself. Microsoft introduced AI channel groups inside Clarity on August 29, 2025, letting the platform separate traffic originating from ChatGPT, Claude, Gemini, Copilot and Perplexity from other referral sources. In December 2025, the company published its research on AI-referred visitor conversion rates. Clarity Citations moved to general availability on May 13, 2026, a release PPC Land covered in detail, giving publishers a structured way to measure how content is discovered and cited inside AI-generated answers, alongside share-of-authority reporting.

That citation infrastructure expanded further at Cannes Lions 2026, when Microsoft bundled several announcements together on June 17, 2026, an event PPC Land reported as combining a set of AI-native grounding APIs called Web IQ, new citation reporting inside Clarity and an expanding Model Context Protocol server for the separate Microsoft Advertising platform, which entered open pilot with read-only access on the same date. That Advertising MCP server is a distinct product from the Clarity MCP Server this article covers; the former connects AI agents to campaign management functions inside Microsoft Advertising, while the latter connects AI tools to behavioral analytics data inside Clarity. Since the Cannes bundle, Microsoft has continued shipping incremental Clarity features roughly every two to three weeks: Topic Insights launched July 9, 2026, adding competitive citation analysis at no additional cost, a release PPC Land examined in the context of rival platforms charging for similar functionality. Query Topics entered beta on July 22, 2026, sorting citation grounding queries into ranked topics, a feature PPC Land reported on alongside its export support for publishers. In early August 2026, Clarity split its AI citation records into branded and non-branded queries, a release PPC Land analyzed for its unresolved methodology questions around brand-token matching. Most recently, on August 13, 2026, Clarity added an AI Scrape-to-Referral Ratio card with operator-level breakdowns, a feature PPC Land covered as the point at which Clarity's citation measurement stack connected crawl activity all the way through to conversion outcomes.

Set against that rapid cadence of citation and AI-visibility features, the August 18 MCP query post stands out for what it does not do. It does not expand the MCP Server's data access, does not raise its request limits and does not connect it to the citation or AI-referral reporting Microsoft has spent the past year building. It is, instead, a documentation and enablement post aimed at helping existing users of the MCP Server extract more value from a tool that has otherwise seen little public iteration since its June 2025 launch, even as adjacent parts of the Clarity product have expanded substantially.

Why this matters for marketers

For agencies and in-house teams already running the Clarity MCP Server, the practical value of this post is largely structural rather than technical. It does not unlock new capability, but it does lower the barrier to writing effective prompts against a system whose underlying interface, natural language, is inherently less discoverable than a menu of dashboard filters. Without an example demonstrating the pattern first, a marketer who has never used the MCP Server has no obvious way to know that combining three behavioral signals into one filter is possible, or that session recordings can be fetched conditionally based on a metrics threshold.

The compound-query approach the post advocates also reflects a broader shift the digital advertising trade press has tracked across nearly every major platform's MCP rollout over the past year. PPC Land's reporting on Adform's 29-skill MCP release in July 2026 noted that platform after platform has followed a similar trajectory: exploration, an initial read-only release, and then a period of skill or query-pattern expansion once the underlying server has stabilized. Clarity's own MCP Server, having launched more than a year before this post, sits comparatively further along that trajectory than servers Google, Meta and others have released more recently.

The distinction between illustrative and measured data in this post also carries a practical caution for readers. Because the five worked examples pair a specific, actionable-sounding prompt with a labeled mock table, it would be easy for a reader skimming quickly to treat the sample percentages, an 82 percent scroll depth here, a 30 percent trend decline there, as representative benchmarks worth comparing against. They are not benchmarks. They illustrate output format only, and any team using these query patterns would generate its own figures from its own Clarity data rather than reading the numbers in Microsoft's post as reference points.

Timeline

Summary

Who: Microsoft's Clarity team, publishing under the byline Clarity Staff, addressing marketers, analysts and developers who use the Clarity MCP Server.

What: Publication of five example query patterns, covering scroll depth against conversion, compound negative engagement signals, multidimensional segmentation, month-over-month trend comparison and metrics-linked session recording retrieval, along with seven general prompt-writing principles.

When: August 18, 2026.

Where: Published on the Microsoft Clarity Blog, describing functionality that operates through Clarity's existing MCP Server, first launched June 4, 2025.

Why: The post aims to demonstrate practical, compound query structures for an interface, natural language prompting through an AI assistant, that offers no built-in menu of examples, at a time when Microsoft has spent the preceding fourteen months expanding Clarity's adjacent AI-visibility and citation features considerably faster than it has iterated on the MCP Server itself.