"We got a couple of inquiries from AI recently" is hard to prove from a dashboard. AI referral traffic is structurally shy.
Why AI traffic goes missing
Three things combine to erase it from your reports.
The parameters are different. ChatGPT tags its landing-page links with its own parameter (oppref and similar), not the gclid used in Google advertising. An analytics tool sees an unfamiliar string and has no reason to guess that it came from a chat.
Mobile browsers strip the source. Many mobile browsers no longer pass the Referrer header on navigation, for privacy reasons. The source is gone before it reaches your server.
Not everyone clicks. Some buyers read the AI answer, then go search your company name on Google. The inquiry owes its existence to the AI, but the credit lands in branded organic search.
So "how many inquiries come from AI" cannot be answered by one tool. You assemble it from layers.
Layer 1: The AI channel in GA4
Since 2026, GA4's default channel groups include an AI Assistant group. When a source matches the list Google maintains, the channel is tagged automatically: medium ai-assistant, campaign (ai-assistant).
In GA4, open the acquisition report, set the dimension to Session default channel group, and look for an AI Assistant row: how many sessions, and which landing pages.
Then open the referral report and search for these domains: chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com. Finding them is proof. Not finding them is not proof of absence, because of the two problems above.
Layer 2: The generative AI report in Search Console
Google launched a generative AI performance report in June 2026. As of 31 August 2026 it is available for every site worldwide. It sits under Performance in Search Console, as its own section.
What it gives you: impressions for your pages appearing in AI Overviews and AI Mode, sliced by page, country, device and date.
What it does not give you: no clicks, no click-through rate, no average position, no queries, and no API.
This is easy to misread. Thousands of impressions do not mean thousands of visitors. The right use is to answer a different question: which of my pages does Google consider worth citing? Take the pages with the most impressions and invest more there.
One setting worth knowing: under Search Console settings there is a "Search generative AI" control that decides, site-wide, whether you may appear in those AI features. It does not affect how you rank in ordinary search. Turning it off removes you from AI answers entirely, which for an export business is rarely what you want.
Layer 3: Bing Webmaster Tools is more granular
If you only watch one console, do not skip Bing's.
Its AI Performance report went into public preview in February 2026 — earlier than Google's — and it shows more: specific citations, the queries that triggered them, and page-level citation activity. Google's report does not offer those yet.
That is the practical reason to have your site in Bing Webmaster Tools: it is one of the few free places that tells you which page an AI cited, and for what question.
Layer 4: One field on your form
The first three layers are tools. This one is a question.
Add a dropdown to your inquiry form: "How did you find us?" Options: search engine / AI assistant (ChatGPT, Perplexity, etc.) / marketplace / referral from a colleague / other.
It works across every channel, needs no parameters, costs nothing, and gives you the answer directly. Its weakness is sample size — few people fill it in. But it is the only method that catches the sessions filed as Direct.
Keep it optional. Adding friction to get data reduces the inquiries you were trying to measure.
What each source can and cannot tell you
| Source | Answers | Blind to | |---|---|---| | GA4 · AI Assistant channel | Whether AI sessions were identified, and where they landed | Sessions filed as Direct | | GA4 · referrals | Which AI domains sent visitors | Sessions that lost the referrer | | GSC · generative AI report | Which pages were shown, in which countries | Clicks, CTR, queries | | Bing Webmaster · AI report | Cited pages and the prompts that triggered them | Every other engine | | Form field | Every channel, including the invisible ones | Precision — it relies on the buyer |
How to use this
Do not add these into a single "GEO traffic" number. The definitions do not match, and the total will mislead you.
Read them as layers instead: AI Assistant sessions tell you how much is already identifiable; GSC impressions tell you which pages AI draws on; the form field tells you how buyers describe the journey themselves. Three numbers, watched as trends, beat any single figure.
To see which of your pages AI cites and what triggers it, see GEO · AI search optimization. To find out where your site stands, ask for a free acquisition audit — the report is actionable on its own, with no obligation.
Two related reads: first check whether your site has anything worth citing, in the quotability checklist; and if inquiries exist but are not being caught, in why inquiries dropped after you added WhatsApp.
FAQ
Why does ChatGPT traffic show up as Direct in GA4?
Two reasons. ChatGPT tags its landing-page links with its own parameter (oppref) rather than Google's gclid, so analytics tools have nothing to recognise. And mobile browsers strip the Referrer header for privacy, so the source is lost in transit. Those sessions end up filed under Direct.
Can Google Search Console show clicks from AI answers?
Not yet. The generative AI performance report gives impressions only, broken down by page, country, device and date. No clicks, no click-through rate, no position, no queries. It answers whether you appeared in an AI answer, not whether anyone came.
If I could only do one thing, what should it be?
Add a How did you find us dropdown to your inquiry form. It is imprecise, but it is the only method that covers every channel, including the sessions that show up as Direct.