Analytics and Measurement
AI and agent traffic measurement
See the part of your traffic that arrives with no referrer and no obvious cause.
What does this work cover?
See the part of your traffic that arrives with no referrer and no obvious cause. In practice that means the GA4 AI Assistant channel group, configured rather than left at its default, and then the harder half: user agent and referrer patterns for the agents that identify themselves.
Why this work exists
It is the kind of thing that gets a line in a plan and no time in a calendar.
This is one part of Analytics and Measurement, described separately because it is usually the first thing anybody needs.
The nearest piece of work to it is building a definition of a lead that both sales and marketing accept. Where that is already handled, this moves faster than the estimate; where it is not, the two get done together or neither of them holds.
What good looks like
The useful version ends with a decision somebody can act on this month, not a report somebody files.
The gap between that and where most accounts start is why this is scoped as AI and agent traffic measurement rather than as a line inside something larger. Running the holdout that tells you what attribution will not is the usual second step once it is in place.
When this is the right work
Start here if you recognise any of these:
- Direct traffic grew and nobody can explain it
- You are asked whether AI is sending business and cannot answer
- Assistant referrals are being counted as direct and ignored
If none of those is true, something else is probably the constraint, and the Analytics and Measurement overview maps the alternatives. The usual neighbours are finding the report somebody rebuilds by hand every Monday and making consent and measurement coexist without pretending.
What the work covers
The work itself:
- The GA4 AI Assistant channel group, configured rather than left at its default. Where the documentation and the actual behaviour part company.
- User agent and referrer patterns for the agents that identify themselves, handled before the parts that depend on it rather than alongside them.
- Branded search as the backstop for the activity that leaves no trace at all. Slow, unglamorous, and the reason the rest holds up.
- Reporting that states plainly which part of this is measured and which is inferred. Boring, necessary, and the first casualty when a deadline moves.
How it runs
The shape of the work, stage by stage.
The first pass
The obvious problems get fixed first, in one batch, so the picture clears before the harder judgement calls have to be made against a noisy background.
Where the judgement goes
What is left needs a decision rather than a fix. These get taken with whoever owns the consequence, not resolved quietly in a ticket.
Checking it held
Changes get re-checked after a fortnight, because a surprising number of them do not survive the next deployment or the next platform update.
What each part runs into
The obstacles differ by part and by market:
- The GA4 AI Assistant channel group, configured rather than left at its default: the Sharjah version of this has to deal with a catchment smaller than the ambition on the website
- User agent and referrer patterns for the agents that identify themselves, where a New York engagement contends with an incumbent whose position comes from presence rather than effort
- Branded search as the backstop for the activity that leaves no trace at all – working in Austin this means accounting for a buying pattern that resolves faster than the reporting cycle
- Reporting that states plainly which part of this is measured and which is inferred, and in Singapore the constraint on it is a sector mix that pulls the brief in several directions
The error worth naming
Expecting clean attribution. Some of this is structurally invisible, and a measurement plan that pretends otherwise collapses the first time it is questioned.
It is a cheap mistake to avoid and an expensive one to discover six months in.
How it is measured
Assistant-attributed sessions and revenue, plus branded search volume year on year. Agreed up front and reported every period, with the raw data available rather than only a summary.
Where the figure is estimated rather than counted, it says so on the report.
What you end up with
At the end of the engagement there is a configured channel view, a baseline, and a monthly read on assistant-sourced activity.
Written to be read by somebody who was not in the meetings.
How it differs by sector
By sector, the same scope produces different priorities:
- Ecommerce and retail: Margin per order decides everything, and the product feed is usually the weakest asset. so this work has to account for that before anything else.
- Manufacturing and industrial, where low volume, high value, and a specification sheet that decides the deal. which moves this work earlier in the plan.
- In B2B services the binding constraint is that a buying committee, a long cycle, and attribution that has to survive both.
The industry pages carry the rest of that detail. Ecommerce and retail and Manufacturing and industrial are the two where this piece of work most often decides the outcome.
Common questions
Can we track AI referrals properly?
Partly. The assistants that pass a referrer can be grouped reliably; agents acting on a user’s behalf often cannot, and information agents that push updates generate no session at all.
Why does branded search matter here?
Because somebody who sees you named in an answer frequently searches your name rather than clicking. That search is the only trace left. GEO or SEO explains the signal. Google Ads or Meta Ads explains the noise around it.
What if we disagree with the diagnosis?
Then we look at the data together. Assistant-attributed sessions and revenue, plus branded search volume year on year is the test, and a finding that cannot survive being questioned is not a finding worth acting on. People arriving here usually end up on How do we measure SEO for a long sales cycle?, and often on Do we need separate pages for every city we serve? after that.
Does anything need doing before AI and agent traffic measurement?
Sometimes. The signals above are the honest test, and where something upstream is the real constraint we will name it even though it is not the work on this page. The sequencing matters: Featured snippet before font-display, not the other way round.
Is there a minimum worth doing?
There is a point below which the measurement cannot detect a change, and we would rather say so than take the work. Assistant-attributed sessions and revenue, plus branded search volume year on year has to be able to move for the engagement to mean anything. Worth a look before committing budget: Brand and Creative for Healthcare and clinics, then Content and Editorial for Healthcare and clinics.
Ask about AI and agent traffic measurement
Tell us what you have tried already. That is usually the fastest route to a useful reply.
Put the data in front of someone who reads it
Share the site and the target. We will say what we would check first and what we would leave alone.