The way people navigate digital channels has changed, with greater engagement happening with Large Language Models (LLMs) and AI agents, alongside humans. To give its customers better visibility into all interactions (human and AI) across the entire customer journey, Contentsquare has introduced new analytics and AI agent capabilities to its digital analytics platform.
Piecing together the customer journey is hard at the best of times, but throw in a search in an LLM or a chat conversation on the website, and things get tricky fast. This is a challenge that Contentsquare wants to help its customers resolve. And it's doing so with some new analytics capabilities.
According to Rachel Obstler, Contentsquare's Chief Product Officer (CPO), the firm is getting many requests for visibility into LLM traffic and how it contributes to the customer journey. Contentsquare provides an out-of-the-box dashboard that shows the percentage of traffic from different LLMs and other sources. You can then follow through from each traffic source and understand how each one is converting. But what if a customer is interacting with you from multiple channels for the same purpose? How do you see all these interactions working across the customer journey, including traditional blind spots like chat? Obstler says:
A really great example of a blind spot is that they know the behavior of the customer on the website, but then they have a built-in chat on their website, and the customer starts a chat. They don't have the same visibility into what's happening inside that chat. Or if the customer decides to make a phone call to support. A lot of times, these support conversations have an impact on the lifetime value of the customer.
Customers can now analyze conversations within conversational chat experiences on the website and combine that data with other customer interaction channels, including support tickets, calls, and in-product chats. Obstler argues that as more websites become conversational, full interactions are happening within these chat experiences, it's important to understand whether customers are talking to AI agents, humans, or a chatbot, and how they view the conversation.
This new analytics capability stems from its 2025 acquisition of Loris, a conversation intelligence platform. It can surface sentiment, intent, and quality signals, providing a deeper view of conversations to identify performance patterns, catch issues, and predict trends, she explains:
Typically, there's this wall or disconnect between the e-commerce team, the marketing team, and the support team. They have different tool sets. They have different organizations, and they don't put this data together. And even if they do put the data together, it's weeks or months later to find out how things are impacting each other. So you can have a product issue that's causing people to call support or contact support and not really understand that's the driver.
For companies that built an app in ChatGPT, they can now measure activity in that app, including the prompt that started it and what customers do inside it. Keep in mind that it's only the prompt that kicked off the app, and not all prompts in the LLM. OpenAI only announced its app capabilities about six months ago, so they are still in their early stages, but there are some good ones, such as Canva, Expedia, Figma, OpenTable, and more. While it may still be a while before they hit mainstream (if they do at all), having the ability available to analyze their use is important for an analytics platform like Contentsquare. Obstler says:
We view that our role is to help our customers understand the experience people are having with their brand. That's why we start with, where's the traffic coming from? But I agree. I think there's probably going to be a time when it's going to be important to understand better some of these prompts and the stuff that's happening in that other interface. But it's the same thing as Google search. You don't really know what's happening in Google because they don't tell you what keywords or things you search on, either. So how do you know how to improve your site, your content, or even your support experience? You need the information to help you figure that out.
The new Sense Agent is helping Contentsquare customers get answers to their analytics questions more quickly and easily. Obstler said it's like having your own personal analyst. You can ask a broad question, and the agent will reason out a plan, perform the analysis, synthesize the data, and provide actionable recommendations. You have full insight into the agent's work to get the answer.
Obstler argues that many support conversations are becoming more commercially-oriented, and as the website becomes more chat-first, there will be more of these conversations. When this happens, responsibility for these conversations shifts away from support to business or marketing, as they affect revenue figures:
The other thing that is making that shift possible is the fact that more and more of these support or commerce-oriented conversations are getting done agentically. So, if they were all still handled by staff, it would make more sense to stay within the support world that manages the staff. As more and more becomes automated, it makes much more sense and makes it possible to have that information freely shared and maybe owned by another team?
Contentsquare wants to democratize access to analytics, and it's doing so by making platform data available to customers wherever they work with a Model Context Protocol (MCP) server. An MCP server is a system that enables external tools and applications to securely access and query data or functionality on a platform in real time. So, you can be in Slack, Claude, Cursor, or ChatGPT and connect to the MCP server to get immediate access to your Contentsquare data.
When would you want to use the MCP server? Obstler says most use cases involve combining two data sets, or asking a quick question in Slack to get an answer. For use cases that require deep analysis or complex questions, it makes more sense to use the Contentsquare platform. But there will also be overlap:
Let's say that you had an error that you know someone just deployed a fix to, or you think someone might have deployed a fix for. You have an MCP server for JIRA and an MCP server for Contentsquare, and you can easily query it and say, 'JIRA, when did we deploy this fix?" And then, Contentsquare, can you tell me how many errors there were before this time and after this time, to very quickly figure out, did it work?'
We're offering that capability as part of the platform, so we're not charging extra for it. I think that there's a big potential for agent-to-agent use cases, and that is a different animal, because the frequency and the volume of that is very different. But when you're talking about a person querying, we think of it as like you could be querying inside our platform using AI, you could be using our UI, or you could be querying from anywhere outside the platform using MCP. To us, it's all the same.
The customer journey is becoming increasingly complex as customers use more channels to find and engage with brands. Contentsquare's ability to connect LLM traffic, conversational data, and support interactions into a single analytical view gives brands a much better understanding of the customer journey and connects marketing, support, and the business in ways other analytics platforms still can't. But there's more to do.
It's still more of a guessing game as to what customers are searching for on LLMs that drive them to a brand's website. LLMs are black boxes, and Obstler said it will be interesting to see how OpenAI, Anthropic, and other AI companies handle prompts in the future.
For now, brands need to get their analytics in order for the data they do have access to. Without a connected view of the customer journey, it's hit-or-miss with the content, messaging, and offers brands provide.
As for the new AI agent and MCP server, these make perfect sense for an analytics platform that wants to democratize analytics and, in fact, are table stakes. Do we still need data scientists and expert analysts? Sure. But by giving more people access to analytics data, more insights become available faster, and better decisions result.
