Microsoft Clarity adds agent-activity reporting and task simulations

What changed

The Agent Activity dashboard shows detected agent traffic over time, the operators visiting a site and the actions they take there. Microsoft names Google, OpenAI and Anthropic as supported operators. The coverage is therefore narrower than every possible automated visitor, and a dashboard total should not be read as a complete census of all AI use.

The accompanying simulation assessments test whether agents can complete important website tasks. Microsoft describes reporting task completion, time taken, number of interactions and opportunities to improve the experience. Initial scenarios focus on commerce and travel, with broader journeys planned. These are simulated task checks, distinct from observing a real customer’s completed purchase.

For advertisers, the useful question is what happens after an assistant reaches a landing page. A product can be discoverable yet difficult for an agent to select in the correct size or take through checkout. Comparing observed actions with a simulated task can help a team identify where to investigate. A successful simulation alone does not prove that an ad caused a sale or that every agent will complete the same journey.

Microsoft’s announcement opens with global availability language, but its individual capabilities have different qualifications. In particular, Ask Clarity, the conversational analysis experience described in the same post, is explicitly a closed beta for retailers. Its planned expansion to advertising data is future work. Those statements should not be combined into a claim that every announced feature is available to every account.

Start by checking which agent views and simulations are present in the relevant Clarity project. Choose a representative task, such as finding a specific variant or completing a booking, and compare any flagged friction with the live site. Keep actual visitor sessions and simulation results distinct in reporting, and verify the available controls and limits before designing a recurring test programme.

Nabil Naghdy’s signed Microsoft announcement is dated 4 November 2026. This article concerns agent-activity measurement and task simulations; it does not present a new paid-ad placement or a promised conversion uplift.