NEWS · OCTOBER 31, 2026 · AGENTIC AI

Microsoft Clarity finds checkout is the most common action in 5.2 million agentic sessions

In a study published on October 29, 2026, Microsoft Clarity analyzed more than 5.2 million agentic sessions across 230,511 projects. Checkout was the most frequently recorded action in agentic sessions (25.21%), against 14.65% in human sessions. Clarity also made its Agent Activity Dashboard generally available to projects that receive agentic traffic.

01 · WHAT HAPPENED?

Clarity measured what agents do once they reach a site

In a blog post dated October 29, 2026 and signed Clarity Staff, Microsoft Clarity published a study of what AI agents do on websites. The starting point was Clarity's own data, where agentic session volume grew 88% in three weeks. In the team's view, knowing that traffic is growing helps, but the more valuable insight is what happens after those sessions land on a site: was a form sent, was checkout reached?

The sample covers August 13 to September 9, 2026 (UTC) and 230,511 Microsoft Clarity projects, more than 5.2 million sessions in total. Agentic sessions were identified through detected agentic browser providers in the OpenAI, Google and Anthropic ecosystems. Human sessions came from the same projects over the same period. The comparison rests on what Clarity calls Smart Events, business-oriented actions such as checkout, form submission, appointment booking, search, download and login.

02 · DETAILS

Checkout leads for agents, outbound clicks lead for humans

Checkout is first in agentic sessions: 306,621 sessions, or 25.21% of them. Form submission follows (241,391 sessions, 19.85%), then outbound click (173,655, 14.28%), appointment booking (102,401, 8.42%), search (87,096, 7.16%) and login (83,096, 6.83%). Further down are "show more" at 52,456 sessions, availability checks at 29,274, add to cart at 28,289 and download at 20,593.

The human ranking looks different: outbound click leads at 26.99%, followed by form submission at 18.56%, checkout at 14.65% and login at 8.84%. So form submission is close across the two groups, checkout is clearly higher for agents, and search weighs more for agents too (7.16% versus 4.79%). Outbound clicks are nearly twice as prominent for humans. Clarity concludes that the groups do not perform entirely different actions; the same actions simply carry very different weight. One reading note: the percentages divide by total sessions, so each shows the share of sessions containing an action, not a share of all actions.

03 · WHY IT MATTERS

Agentic sessions do more than browse, they reach transaction points

Clarity says the numbers show that sessions involving agents are not only passive browsing or information gathering; they also reach commercial, task-oriented interactions such as checkout, booking and forms. The post adds that understanding why these actions show up more often when agents are involved, and how to accommodate agents taking them, will matter as agentic commerce develops. Telling agentic traffic apart is framed as a separate measurement question.

The study's limits are stated plainly, and reading the news correctly depends on them. A session classified as agentic may include a human handoff, and Clarity cannot currently split the agent and human portions into separate sessions. The findings therefore describe activity recorded in sessions involving agents, not actions an agent completed on its own. Smart Events are independent session counts, not sequential funnel steps or verified task completion. The study also covers sites that use Clarity; that is our inference, but it means the results should not be generalized to the whole web. We covered Clarity's AI scrape-to-visit ratio metric earlier.

04 · TÜRKİYE

The source says nothing about Türkiye, so this section is commentary

Nothing in the Clarity post concerns Türkiye: no project or site from Türkiye, no agentic session traced to Türkiye, no local date or institution. This section is therefore not a fact drawn from the source but plainly UNALSOFT commentary.

Our reading: the results come from sites that use Clarity and cannot be transferred directly to sites in Türkiye. They still point in a direction: a site with checkout, form and booking flows may meet agents as visitors. The questions we cannot answer are these: what share of sessions on Turkish sites involve agents, which sectors they concentrate in, and whether agents complete these actions alone or with a human handoff. You can only see that by separating agentic sessions in your own site's analytics.

The UNALSOFT view

We read this as a matter of measurement and flow hygiene. If sessions involving agents reach checkout and form steps, then for e-commerce and appointment-driven small business sites it pays to keep those flows clear, predictable and error-free for human and agent visitors alike. A second practical step is tracking agentic sessions separately in analytics; otherwise traffic and conversion numbers become harder to read correctly. If you want to review your own store or booking flow with that lens, our e-commerce panel service fits. Because the source has no data on Türkiye, we propose no local conclusion.

Are your checkout and form flows ready for agent visitors too?

A short conversation is enough to review your store or booking flow from both the human and the agent visitor's point of view.

Message on WhatsApp