Microsoft Clarity is extending website analytics to account for visitors that are software agents rather than people. On November 4, the company introduced Agentic Analytics, expanded Brand Agents and detailed Ask Clarity, bringing several parts of the AI-assisted customer journey into its free behavior-analytics platform.
For marketers, the immediate question is practical: can an AI agent find a product, interpret the offer and complete a conversion without getting stuck? Traditional session recordings and conversion funnels were designed around human clicks. Microsoft’s new tools aim to expose automated journeys that those dashboards may otherwise miss, although some capabilities remain in limited access.
Table of contents
Jump to each section:
- How Agentic Analytics follows automated website journeys
- What Brand Agents and Ask Clarity add
- Why marketers need a separate view of agent traffic
- What teams should test before relying on these metrics
Key Takeaways
- Agentic Analytics identifies supported AI operators and offers tests for website tasks such as checkout.
- Brand Agents addresses product discovery and local-service interactions, while Ask Clarity provides conversational analysis of session data.
- Marketers should separate observed agent behavior from verified customer outcomes and check which features are available to their business.
How Agentic Analytics follows automated website journeys
Agentic Analytics helps site operators identify visits from supported systems operated by OpenAI, Google and Anthropic. Microsoft’s Agent Activity dashboard is intended to show which agents arrive, what tasks they attempt and how their activity changes over time. Examples include searches, form submissions, logins, downloads, bookings and checkout interactions.
The distinction matters because a successful page load is not the same as a successful customer transaction. An agent might recognize a product but miss a discount, fail to select a delivery option or encounter a form that depends on a human action. These are experience failures that aggregate traffic charts may hide.
Microsoft is also adding simulation assessments for selected agent workflows. Teams can test whether a task completes, how long it takes and where the interaction breaks. The initial simulation emphasis is commerce and travel, with other journey categories planned. That scope should prevent an early buyer from assuming every type of customer action is supported today.
5.23 million agentic sessions across 230,511 projects formed the August 13 to September 9, 2026 observational sample reported by Microsoft Clarity. It is a product-specific dataset, not an estimate of all AI-agent traffic on the web.
The company has separately described rapid growth in detected AI-agent activity. That trend supports looking at automated visits as a distinct analytics problem, but growth in detected sessions alone does not establish additional sales or incremental demand. Marketers still need an outcome measure for the conversion they care about.
What Brand Agents and Ask Clarity add
Brand Agents tackles a different part of the journey. Microsoft is connecting branded product experiences with discovery in Copilot, Bing and MSN, while extending a website assistant to local businesses that want to answer questions, capture leads and help customers book services. This is primarily a customer-interaction and discovery feature, rather than a neutral measurement layer.
Ask Clarity is closer to an analyst’s interface. Instead of manually filtering recordings and heatmaps, teams can ask natural-language questions about user behavior and receive responses grounded in session evidence. Microsoft’s initial focus is behavior analytics, with AI visibility and advertising data described as future areas of expansion.
Availability is an important qualification. Microsoft describes the overall Clarity expansion as globally available, but its blog also identifies a closed beta and waitlist for retailer access to Ask Clarity. The precise eligibility and configuration of individual experiences therefore matter more than the broad launch headline. Organizations should verify what their own account can activate before promising a rollout.
The product announcement includes a user example from Kelly Wellman, ecommerce manager at Pela Case, who said, “I use this feature to dive deeper into user behavior.” She described using it to investigate where shoppers stopped interacting on a product-personalization landing page. That is useful qualitative evidence of the workflow, though it does not establish typical results for all retailers.
Why marketers need a separate view of agent traffic
Website measurement has generally been organized around a human visitor progressing from a campaign to a landing page and finally to a conversion. Agent-assisted buying complicates the sequence. An AI service may recommend the product without delivering a conventional referral, or browse and interact with the merchant site on the customer’s behalf.
This creates two separate marketing questions. The first is whether the brand was discoverable in AI answers and agent recommendations. The second is whether the website could actually serve the agent after discovery. Brand Agents targets some of the former and customer interaction; Agentic Analytics is meant to expose the latter. Neither on its own proves how much revenue would not have occurred otherwise.
Clarity also sits in a competitive analytics market that includes Google Analytics for event and conversion reporting, and behavioral-analysis vendors such as Hotjar and FullStory for session-level understanding. Microsoft’s differentiation in this announcement is the effort to make supported automated-agent actions visible alongside human behavior. Integrations, identification coverage, error rates and operational usability will determine whether that additional view becomes part of routine performance reporting.
Industry coverage from Search Engine Land similarly identifies agent activity, Brand Agents and natural-language behavior queries as the three major changes. Buyers should treat the announcement as an early capability map, rather than confirmation that agentic commerce attribution is already solved.
What teams should test before relying on these metrics
A sensible first test is a small set of high-intent tasks, such as finding a specific product, applying a promotion, completing an appointment request or reaching the final checkout step. Record what an agent succeeds at, where it stops and whether the same journey works for a human user. That provides a more useful baseline than simply counting automated visits.
Next, segment the data. Compare supported agents separately from crawlers, ordinary visitors and unclassified automated traffic. If an agent cannot be reliably identified, do not silently attribute it to a specific provider. Ask the vendor how detection and task simulations handle unsupported browsers, privacy settings and changes to agent behavior.
Finally, align any dashboard finding with business outcomes. A decline in agent task failures may indicate a more navigable site, but it does not equal incremental bookings or orders. Measure completed conversions, lead quality and the cost of remediation wherever those measures are available. A controlled before-and-after test on one journey is more defensible than promising a broad uplift.
The new Clarity features offer marketers a starting point for understanding how automated visitors encounter their sites. The operational value will come from identifying specific friction, fixing it and confirming whether those changes produce better outcomes for customers.
