Measuring AI search traffic

Everybody wants a number for AI traffic. The honest position is that a slice of it is measurable, a larger slice is not, and the most useful signals are indirect.

What you can see

• Referrals from assistant domains. Visits arriving from the major assistants show up in analytics as ordinary referral traffic. Group them into one channel.
• Behaviour of those visits. They often arrive further down the funnel, so compare their conversion rate with organic rather than judging by volume.
• Query level click through changes. In Search Console, falling click through on question style queries with stable impressions suggests an Overview is absorbing clicks.

What you cannot see

Citations that produce no click. Answers given from training data with no source shown. Whether an assistant recommended you to somebody who then searched your name. This is a real measurement gap and no tool currently closes it credibly.

Proxy measures worth tracking

1. Branded search volume. If people are hearing about you somewhere you cannot see, this is where it surfaces. Filter for it in Search Console.
2. Direct traffic trends. Crude, noisy, and still informative alongside the rest.
3. Manual citation checks. Ask your ten key questions in each assistant quarterly and record who is named.
4. Enquiry source questions. Add "how did you hear about us" to your form and read the answers.

Putting it in reporting

Add an AI referral channel and a branded search line to the monthly report, and leave the rest alone. Rebuilding your entire measurement approach around a channel producing a small share of visits is a mistake, and SEO reports explained sets out what should stay at the centre.

If an agency claims precise AI visibility measurement, ask exactly how it is derived. Confident numbers in an area with no reliable data source belong with the other SEO agency red flags. To compare how different firms approach it, request quotes from Bristol SEO agencies.