We were in a research session a few weeks ago. Healthcare Professionals, observing how they research and choose a service provider. Finding possible options, exploring websites, walking us through the decision-making process. It is the kind of work we have done hundreds of times.
One of the participants opened Copilot. Typed a question, refined it, asked a follow-up. Within a couple of minutes they had a clear summary of the options available to them.
Then they did something that caught my attention.
Instead of clicking one of the links Copilot provided, they opened a new browser tab, typed the provider’s name into Google, and went to the website from there.
Small moment. Easy to miss in the flow of a session. But then the next participant did the same thing. And the next.
Three participants. Three times the same pattern. Not enough to hang our hat on. But enough to make me pay attention.
A behaviour worth paying attention to
I have been thinking about it since.
Not the behaviour itself – that is simple enough. A person uses an AI assistant to research something, finds a company they are interested in, and rather than clicking the link the AI provides, goes to find the website themselves. What I keep coming back to is what happens next. What happens when that person arrives.
Because in the analytics, they look like any other branded search visitor. Someone who typed the company name into Google, clicked the top result, and landed on the home page. The data says: straightforward visit, someone who already knows the brand, or maybe someone right at the beginning of their journey.
But that is not what is happening. Not for this person.
This person has already done their research. They have already compared. They have already been given a reason to look at this specific company by an AI, in a conversation the organisation knows nothing about.
They are not arriving to browse. They are arriving to check.
Two visitors, same front door
Think about two people who both land on the same website, via the same route – a Google search for the company name.
The first is genuinely starting out. Heard the name somewhere. Curious. They want to understand what this company does, whether it is relevant to them, how it might help. The home page, at its best, is designed for this person.
The second has just spent five minutes with an AI. They have explored the landscape, compared options, and arrived with a reason. They are not browsing. They are validating. Does this company match what the AI told them? Do they trust it? Is there enough here to take the next step?
Both visitors typed the same short search term. But they are in very different positions. One needs discovery. The other needs reassurance.
And the website, in all likelihood, treats them identically.
You cannot see this in the data
There has been a lot of effort to track AI referral traffic over the past year. Some AI platforms now pass referral data when a user clicks a link inside the AI. ChatGPT appends tracking parameters to some of its citation links. Google has introduced ways to identify AI-sourced visits in GA4.
But that is tracking one thing: clicks from within the AI.
What I observed is different. These users did not click the AI’s link. They opened a separate browser and typed the company name into Google. There is no referral tag. No UTM parameter. No cookie trail connecting the AI conversation to the website visit. From an analytics perspective, the journey does not exist.
No amount of configuration will change that. The data cannot show you something the user deliberately bypassed.
Where the focus seems to be
There is a lot of conversation at the moment about AI visibility. How to make sure your brand appears in AI-generated responses. How to optimise for large language models. How to position your content so that AI mentions you.
Fair enough. Being mentioned matters.
But it is only half the question.
If more people are being directed to your website by AI – arriving with context and expectations already shaped – then what they find when they get there matters just as much as whether they were sent in the first place.
Visibility gets them to the door. What is behind it determines whether they stay.
The risk is that organisations invest heavily in being mentioned by AI while the experience those visitors actually encounter has not changed at all. The same front door, designed for a visitor who may no longer be the typical one walking through it.
What about your website?
Think about your own home page for a moment. How many of the people landing on it today typed your name into Google after a conversation with an AI? Unfortunately, you are unlikely to know. Your analytics will show a branded search, and that is all they will show.
Are more and more visitors arriving AI-primed – already informed, already comparing, already expecting something specific? How do you cater to this user base when you cannot tell them apart?
I noticed this because we were in one-to-one sessions, watching real people research in real time. It is not the kind of thing that shows up in a dashboard or a quarterly report. And whilst we are starting to see the behaviour more, the work is in further understanding these user expectations and reshaping the landing pages to accommodate this new journey.
I will keep investigating this and it will likely dominate my conversations for the next month or two. If it raises questions about your own visitors, I would be glad to hear from you.
Ali Carmichael is the Founder and Managing Director of Experience UX, a UX research and user-centred design consultancy.
More reading the will interest you: The Outside-In Paradigm: Why Smart Organisations Still Get It Wrong
