Spoke
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.
A practitioner method. Its known biases are well established in survey research; no platform documents or endorses it, and none of the figures it produces are presented here as benchmarks.
Every sophisticated measurement approach on this site works around a hole: exposure that produces no session. The crude approach walks straight into it and asks.
Is "how did you hear about us" worth asking?
Yes, and it is the single highest-value addition to most measurement setups, precisely because it captures what analytics structurally cannot. A person who read an AI answer, remembered your name and searched for you directly appears in analytics as direct traffic. Asked, they will tell you where they actually heard of you.
It is biased and imprecise. It is also first-hand testimony about the part of the funnel no instrument can see, which makes it worth the one field it costs.
Designing the question
Make it required, and ask at first contact. Asked later, you get the last thing they remember rather than the first thing that worked.
Offer specific options, not an open text box. Free text produces unanalyzable variety. A short list with an "other" option produces data.
Include an AI assistant option explicitly. If the list does not name it, people pick "search engine" or "online" and the signal disappears into a bucket you already had. This is the single most important detail on this page.
Keep the list short. Seven or eight options. A long list produces selection of whichever is first.
Allow more than one where it fits. Real discovery is usually several touches, and forcing a single choice discards that.
Known biases, named
So you read the results correctly rather than trusting them:
Recency. People name the most recent touch, not the first.
Prominence. Memorable channels get over-reported. A billboard is remembered; a search result is not.
Social desirability. Some people answer with what sounds sensible rather than what happened.
Recall decay. Anything more than a few weeks back is reconstructed, not remembered.
None of these is fixable. All of them are roughly constant, which is what makes period-over-period comparison meaningful even when the absolute numbers are not.
Reading it properly
Track the mix over time, not the raw counts. If the share naming an AI assistant rises across three quarters while your panel share also rises, those two independent instruments agreeing is the strongest evidence available in this field.
Cross it with outcomes. Which self-reported source produces the best-qualified contacts? That question connects marketing to the business, and no analytics platform can answer it because it requires the outcome, which lives in your own records.
Why it belongs in the operating rhythm
The hard part is not the field. It is that somebody has to ask it consistently, at first contact, every time — which makes it an intake discipline problem rather than a marketing one.
That is the same conclusion the missed-calls spoke reaches from the advertising side. The instruments that measure modern search visibility mostly live in operations, which is why the measurement owner and the intake owner need to be looking at the same report.
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.
The 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.
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.
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.