Chatbot Users May Be “Super Intentional.” So Why Is Intent Getting Harder to Measure?
- Jul 1
- 3 min read
A few weeks ago, while auditing engagement data for a client campaign, the numbers looked solid at first glance. But a closer look at the traffic patterns revealed something worth examining. It wasn’t fraud in the traditional sense. It was legitimate-looking activity that the ad platform had learned from and optimized against.
Around the same time, we were doing what a lot of people now do: using an AI chatbot to research a product. In this case, it was a robot vacuum. Not a quick Google search, but a back-and-forth conversation about whether a wet/dry robot vacuum made sense for our home or whether a standard stick vacuum was the smarter call. By the end, we had visited several brand sites, compared models, evaluated price points, returned to the chat window a few more times, and eventually made a purchase.
That raised a different question: what would those website sessions look like in a brand's analytics?
One visit, maybe two. Solid time on site. No conversion yet. To the brand, it likely looked like a high-intent shopper in research mode.
But was it really? We were still exploring, not necessarily ready to buy. The chatbot was doing much of the filtering for us, and the brand-site visits were only one part of that process. If those signals flowed into optimization systems and were interpreted as meaningful human intent, the issue was not whether the traffic was real. It was whether it represented what brands thought it represented.
That’s a different problem.
Why This Matters Now
AI systems are increasingly active across the open web. AI crawlers generate large-scale traffic. Ad buying platforms use AI to define and optimize performance. Synthetic inputs are entering testing and targeting workflows.
None of this is inherently problematic. But it introduces a dynamic that most measurement frameworks weren’t designed to handle: Not all non-human activity is malicious, but it can still look like signal.
The issue extends beyond traffic quality. As conversational AI becomes part of the research process, brands increasingly see only fragments of the customer journey. A user may spend twenty minutes exploring options with a chatbot, then visit a handful of websites that appear highly qualified in analytics despite the fact they're still early in their decision-making process.
OpenAI recently described chatbot users as "super intentional," people arriving with a specific job to be done. That may be true. But intentional doesn't always mean transactional. Someone researching a future vacation, comparing software vendors, or evaluating a robot vacuum can generate many of the same signals as a high-intent prospect while still being firmly in exploration mode.
The signal looks the same. The intent behind it may not be.
The challenge isn't whether the traffic is real. It's whether the signals being fed into optimization systems represent the kind of intent brands assume they do.
The Real Risk
Modern ad buying systems don't just measure behavior. They optimize against it. And optimization systems rarely understand intent directly. They learn from observable behaviors that are assumed to correlate with it.
Performance may remain stable on paper while quietly drifting from actual business impact.
The Question Has Changed
Historically, marketers asked whether traffic was real or fake. Increasingly, the more important question may be whether that activity represents the kind of human intent brands think it does.
That’s a harder question, and one most ad buying platform don’t or can’t explicitly answer. It connects directly to our earlier post on Platform Metrics vs. Business Reality. OpenAI’s own Cannes comments suggest even the newest entrants are still working through the same measurement gap.
The Open Questions
It’s not clear how material this is today. But the direction of travel is worth watching:
How much of the data informing optimization is truly representative of human purchase intent?
Are measurement frameworks built to handle traffic that is real but not ready to buy?
Who ultimately defines what “performance” means: the brand, or the platform’s algorithm?
Bottom Line
This isn’t about AI causing bot traffic. It’s about a more important measurement challenge.
As AI becomes a research tool, an ad surface, and an optimization engine, brands will need to look beyond whether traffic is real and ask whether the signal is meaningful.
The last decade's measurement challenge was filtering bad traffic. The next one may be interpreting real traffic whose intent is increasingly difficult to see.
— Dugbe
Disclaimer: This post was refined with the help of AI tools for clarity and structure. The thinking and perspective are entirely ours. We use tools thoughtfully. We still think for ourselves.
