Spoke
The zero-click attribution gap
Being cited without being clicked is a real outcome that last-click reporting cannot see. It usually resurfaces later as direct or branded traffic, credited to the wrong thing.
That generated answers produce impressions without clicks is documented by Google's own reporting, which publishes impressions and not clicks for those features. The downstream attribution behavior described here is reasoning from how last-click models work.
A firm becomes more visible, gets cited in answers people act on, and the dashboard shows informational traffic falling. Both things are true at once, and the reporting is built to show only one of them.
Why is AI search traffic so hard to attribute?
Because the valuable outcome frequently produces no visit. Someone asks a question, your page is cited, they read the answer and act on it without clicking. There is no session to attribute, and Google's generative AI reporting reflects this directly — it reports impressions for those features and not clicks.
When that person does arrive, it is often days later as a direct visit or a branded search, which last-click attribution credits to "direct" or "brand" rather than to the exposure that caused it.
The sequence
- Somebody asks an assistant a question in your field.
- Your page is retrieved and cited.
- They read the answer. No click.
- They remember the name, or they do not.
- Weeks later they search your firm by name and call.
Last-click records step five. Steps one to four produced it and are invisible. Run that across a quarter and a real channel is doing real work while the report shows a decline in the informational traffic it replaced.
Why you cannot fix it
This is worth saying plainly, because a lot of money is spent trying.
There is no session, so there is no referrer, no landing page, no campaign parameter. The event happened entirely inside somebody else's interface. No tag, no server-side configuration and no attribution model recovers data that was never generated.
The engines that do tag referrals only tag the minority who clicked. That is useful and it is not the gap.
What to do instead
Accept the shape of the evidence. You are going to triangulate from several partial sources rather than measure one number. Three partial views, held honestly, beat one precise number that is measuring the wrong thing.
Ask people. The self-reported attribution field is crude, biased by memory, and the only place a human tells you directly that a machine recommended you.
Watch the leading indicators. Branded search volume, direct traffic and first-contact quality all move when AI exposure is working, lagging and confounded but directionally real.
Track share of voice separately. The prompt panel measures exposure without needing a click, which is precisely the thing sessions cannot see.
Change the headline metric. Revenue per session rises when fewer, better-qualified people arrive. Sessions falls. Reporting the second as the top line tells a story that gets less true each quarter.
The honest framing for a report
Say the quiet part in the report rather than letting someone discover it:
Informational sessions are down. Qualified contacts are flat or up. Self-reported source shows an increasing share naming an AI assistant. We believe the channel is working and we cannot prove it with session data, because the mechanism does not produce sessions.
That paragraph is more defensible than any model that assigns a confident number, and it survives the question "how do you know?" — which the confident number does not.
More in Measurement
Prompt panel tracking
Asking an assistant about yourself once tells you nothing. A fixed set of prompts, re-run on a schedule and recorded, tells you a direction — which is the most that is available.
AI visibility reporting
Google and Microsoft now report AI visibility directly, and OpenAI tags its referrals. That beats a third-party citation tracker — and each one measures something narrower than its name suggests.
AI share of voice
How often you are named across a fixed set of questions, relative to the people you compete with. It is noisy, it is the closest thing available, and it breaks the moment anyone reports it as a single number.
AI referrers
The visitors who do click from an AI answer are identifiable in ordinary analytics. It is the easiest measurement win available here, and it counts the smallest part of the channel.
Self-reported attribution
One required field on every intake. It is biased, imprecise, and the only place a human tells you directly that a machine recommended you.
KPIs that survive
Sessions as a headline metric tells a story that gets less true every quarter. These are the measures that keep meaning what they meant when the traffic mix changes underneath them.