Featured · client work
Lunarlink Solutions, Toronto
See it live
Traffic from AI assistants arrives as ordinary referral traffic, so it is tracked by segmenting referrers such as chatgpt.com, perplexity.ai, claude.ai, and gemini.google.com in analytics. That number counts only the visitors who clicked through. It cannot see the larger group who read the answer and never left the assistant.
In GA4 this is a custom channel group, or a comparison filtered on session source, matching the assistant hostnames. Give it its own name so it stops being swept into Referral alongside directory listings and forum links.
Treat the resulting number as a floor rather than a count. Some assistants strip the referrer, some open links in an in-app browser that reports as direct, and a link copied out of a chat and pasted into a browser carries no referrer at all. The trend across months is the useful part, not the absolute figure in any one month.
Search Console reports Google Search. Clicks from an AI Overview are folded into ordinary Search totals with no separate label, so a page can lose clicks to an Overview that quoted it and the report shows only that the click-through rate fell.
Nothing in Search Console reports ChatGPT, Perplexity, Claude, or Copilot at all. Any measurement of those has to come from analytics on one side and direct testing on the other.
Write twenty questions the way a customer would type them, covering the services and the cities that matter. Run the same twenty, unchanged, on the same day each month across the two or three assistants worth caring about. Record whether the business is named, and copy out the passage that was quoted.
The quoted passage is the valuable half. It tells you which page shape is being retrieved, which is far more actionable than a count of appearances, and it usually shows that short, direct pages are being pulled while the long marketing pages are not.
Very little for the first quarter. Retrieval indexes and training data both lag the web by months, and a page published in September may not be quotable until well into the following year.
When the sessions do appear they tend to be few and unusually far along, because the visitor arrives carrying a recommendation rather than a list of ten options. Judge the channel on what those visitors do after landing, not on the session count, which will look negligible next to organic search for a long time.
| Surface | Visible in Search Console | Visible in analytics |
|---|---|---|
| Google AI Overviews | Folded into Search totals with no separate label | No. The click is recorded as Google organic |
| ChatGPT | No | Yes, as chatgpt.com, when the visitor clicks through |
| Perplexity | No | Yes, as perplexity.ai |
| Claude | No | Yes, as claude.ai |
| An answer read without a click | No | No. Nothing on the site records it |
Send the link. We read the site by hand and write back inside five business days.
No call, no commitment
