Contentsquare now tracks AI agents as their own kind of website visitor

That sounds like a reporting tweak. For whoever owns the conversion rate at a retailer or consumer brand, it changes what the number means. The catch: the reworked platform is in beta with selected customers, ContentGrip reported, with broader access to follow in phases.

Humans, LLM referrals and agents in one record

With the update, Contentsquare splits out LLM and AI agent traffic and analyzes it next to human sessions. The aim is to help brands tune journeys for LLM referrals and for agentic commerce, rather than for human visitors alone. CX Foundation describes agentic commerce as customers asking a large language model to research, compare, decide and, in some cases, complete purchases on their behalf.

Around that split, Contentsquare layered several capabilities. Conversation intelligence pulls voice and text chats, chatbot exchanges and agent dialogue into the journey as events, then links them to friction on the site or in the app. Multi-session analysis follows one customer across sessions, channels and devices. Automated WebMCP monitoring captures autonomous agent activity that, the company says, most tools miss entirely. A multi-touch attribution model then credits conversions across every touchpoint, AI-referred ones included, rather than handing it all to the last click.

CX Foundation's Charlie Mitchell, who covered the CX Circle launch, offered an illustration of why the stitching matters. A shopper asks a personal agent to suggest new shoes, follows the link it surfaces, opens a chat to ask about sizing, scrolls back through the product page and leaves. Seen separately, that's four unrelated events in four systems. Mitchell reported that the platform is designed to connect the AI referral, the chat, the page behavior and the abandoned cart, then recommend changes to the site and the conversation design.

Feliz Papich, SVP of digital technology, experience and insights at footwear brand Crocs, said in the launch release that Crocs customers move among the app, the website, a chatbot and sometimes an AI agent the brand can't see directly, and that Contentsquare ties those threads together and points to what to act on.

What a mislabeled visit does to a funnel

ContentGrip, in its analysis of the launch, set out the measurement problem plainly. An agent arriving at an ecommerce site may be carrying out a legitimate request from a real shopper. Blend those visits into human sessions and bounce rates, conversion funnels and retargeting audiences all skew. Throw every automated visit out as bot traffic and a new sales channel disappears from view.

Once an AI agent can run a shopper's errand on a retail site, the label on that visit decides what the bounce rate, the conversion funnel and the retargeting audience actually measure.

Attribution needs the same care. ContentGrip noted that seeing an AI referral in a journey doesn't prove which earlier recommendation caused the sale, and that marketers should check which connections and identity signals their own setup actually captures. For teams whose AI-referred traffic is still a small share of visits, the separation mostly brings cleaner baselines today. For brands where agents already browse and buy, it starts to look like a new channel with its own funnel.

Sense moves from flagging problems to running tests

The second half of the relaunch is about what happens after a problem is found. It now estimates the financial impact, ranks issues by business value, and writes the resulting emails, tickets and notifications straight into a team's existing tools. It also has a new mobile app and adds memory, context and a Visual Stories format.

The bigger change is native agentic experimentation. According to the company, Sense forms hypotheses, designs and codes, launches tests and validates the revenue impact, all in one place, and can run that loop end to end; all a team needs to do is approve. Mitchell laid out the same sequence as a numbered workflow beginning with an identified opportunity and its estimated financial impact. For a team that still moves an idea from analyst to designer to developer to testing tool by hand, that is the part most likely to change how the week is spent.

Benchmarks, the other new capability, compares specific features, page elements and user flows against industry peers and tests whether a difference actually links to better performance. Mitchell's example: for a product page converting below expectations, it points to the weak components and flows instead of offering one site-wide conversion rate to aim at. Brands can also compare how ready their sites are for personal agents against competitors.

Through the Model Context Protocol, Contentsquare also pushes its data into ChatGPT, Claude, Microsoft Copilot, VS Code Copilot, Cursor, Figma, Dust and Amazon QuickSight, among other tools. Teams that already run their own testing stack get the partner integrations instead. ContentGrip cautioned that implementation details and entitlements will decide how much of the work is really automated in a given account.

A beta in a crowded category

The strongest objection to reading too much into this is volume. ContentGrip observed that the commercial scale and timing of autonomous agent interactions still vary widely by sector, so a separate agent funnel may be essential for one retailer and close to empty for another. Contentsquare's announcement doesn't say how many customers are live in the beta or what agent traffic volumes it is processing.

The category is also busy. Amplitude, Fullstory and Quantum Metric cover overlapping ground in product behavior, session analysis and conversion friction, and Amplitude's 2026 annual filing names Contentsquare among its competitors. What Contentsquare is betting on is continuity: one record running from the human journey to the agent journey to the ranked fix to the test that proves it. A feature list can't settle whether that beats a combination of separate analytics and testing tools. Results can.

Those results will show up as the phased rollout moves forward. Until then, the most useful thing a digital analytics lead can take from the launch costs nothing: knowing how today's dashboards already treat the agents showing up on the site.