Agentic Commerce
Agent traffic attribution
See the revenue assistants are sending you, which currently arrives looking like direct traffic. Scoped as its own piece because that is how it is usually bought.
What does agent traffic attribution actually involve?
See the revenue assistants are sending you, which currently arrives looking like direct traffic. In practice that means the GA4 AI Assistant channel group introduced in May 2026, configured rather than left at default, and then the harder half: referrer and user agent pattern work for the agents that identify themselves, and honest handling of the ones that do not.
Why this work exists
Most of the cost here is not the work itself. It is the months spent on the wrong thing because nobody checked this first.
Agentic Commerce is the wider programme; this is the piece that gets scoped first more often than any other.
The nearest piece of work to it is getting the product feed accurate enough for an agent to act on. 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 agent traffic attribution rather than as a line inside something larger. Making stock and shipping truthful in real time rather than nightly is the usual second step once it is in place.
When this is the right work
This is the right work when any of these is true:
- Direct traffic rose and nobody can explain it
- You are being asked whether AI is sending business and have no way to answer
- Conversion rate from an unexplained source looks poor while the customers look unusually good
If none of those is true, something else is probably the constraint, and the Agentic Commerce overview maps the alternatives. The usual neighbours are finding the variants an agent cannot distinguish and fixing the data and deciding whether in-chat checkout is worth building for this catalogue.
What the work covers
The pieces of work inside it:
- The GA4 AI Assistant channel group introduced in May 2026, configured rather than left at default, revisited rather than finished, because the inputs keep moving.
- Referrer and user agent pattern work for the agents that identify themselves, and honest handling of the ones that do not. The effort here scales with the size of the estate, not with the size of the brief.
- New-customer rate as a separate measure, since assistant traffic behaves differently from search traffic once it lands, with the evidence kept so the decision can be defended later.
- A reporting view that survives the fact that some agent activity generates no session at all. This is the piece a client can verify themselves, which matters more than it sounds.
How it runs
It runs in three passes. The order is not interchangeable.
Clearing the contradictions
Where two systems disagree about the same fact, that gets settled before anything is built on top of either of them.
Setting the review rhythm
How often this gets looked at again, and by whom, is agreed now rather than left to whoever notices first.
Costing the options
Where there is more than one route, each gets a rough cost and a rough timeline, so the choice is made on trade-offs rather than on whichever was described most confidently.
What each part runs into
In practice each of those four runs into something specific:
- The GA4 AI Assistant channel group introduced in May 2026, configured rather than left at default – working in Abu Dhabi this means accounting for a reporting line that averages away the market in question
- Referrer and user agent pattern work for the agents that identify themselves, and honest handling of the ones that do not, and in Leeds the constraint on it is a market where buyers qualify by area rather than by city
- New-customer rate as a separate measure, since assistant traffic behaves differently from search traffic once it lands, harder in Dallas because of a competitive set that changes street by street
- A reporting view that survives the fact that some agent activity generates no session at all – in Melbourne, complicated by the fact that a decision made by somebody who never visits the site
The error worth naming
Judging the channel on conversion rate alone. Walmart found in-chat completion converting at about a third of a click-through while the same traffic produced roughly twice the new-customer rate.
The reason it matters is that the failure is invisible from the report: the activity happened, the number did not move, and the conclusion drawn is usually the wrong one.
How it is measured
Assistant-attributed sessions and revenue, and new-customer rate against your site average. The measure is specific on purpose: a headline figure that cannot move inside the engagement is not useful.
The point of the number is to settle an argument, so it is chosen for being hard to dispute rather than for being flattering.
What you end up with
The concrete artefact is a channel configuration, a baseline report, and a measurement approach that states plainly what is counted and what is modelled.
It includes what was looked at and ruled out, which is usually the more useful half.
How it differs by sector
The brief changes with the sector, not just the emphasis:
- Travel and hospitality, where aggregators hold the demand, so direct booking depends on being the specific named property. which moves this work earlier in the plan.
- In SaaS and technology the binding constraint is that product-led companies where the buyer self-educates and the trial is the sales call.
- Finance and fintech: the thing that changes the scope here is that regulated claims, compliance review in the publishing path, and trust signals that have to be real.
The industry pages carry the rest of that detail. Travel and hospitality and SaaS and technology are the two where this piece of work most often decides the outcome.
Common questions
Can agent traffic be tracked accurately?
Partly. The agents that identify themselves can be grouped reliably. The ones that do not, and the information agents that push updates without a session at all, are structurally invisible, which is why branded search matters as a backstop.
Why does this traffic convert differently?
Because the research happened before the click. The visitor arrives further along, which raises intent and lowers the number of pages they need.
Is the work transferable to our team?
That is the intent. A channel configuration, a baseline report, and a measurement approach that states plainly what is counted and what is modelled. Written for somebody who was not in the room, because the alternative is a dependency rather than a result. When this comes up in a pitch, Chandigarh and Chennai are the pages that settle it.
How do we know the finding is real?
Because the measurement is shown rather than summarised. Assistant-attributed sessions and revenue, and new-customer rate against your site average is reported against a baseline taken before anything changed, with the underlying data available. The evidence is in Ecommerce and retail. The argument about what it means is in Education and training.
What comes after this?
Whatever Assistant-attributed sessions and revenue, and new-customer rate against your site average shows. The next piece is scoped from the result rather than agreed in advance, because committing to it now would be guessing. If this is the part you came for, Agentic Commerce goes further into it, and App Store Optimization covers what sits either side.
Ask about agent traffic attribution
Write what you would say on a call. The reply will be the same thing we would have said.
Have someone senior look at the numbers
We would rather tell you the work is not worth the money than take it and report on activity.