AI Assistant Referral Traffic in GA4 for Local Businesses

AI assistant referral traffic in GA4 is a partial view: you will never see all of it. I'm Marven Salgado, founder of Social Influence in Mascouche. Québec owners may suspect people are finding them through ChatGPT, Gemini and other AI assistants, but their GA4 reports may not show it clearly. This is where to look and what to check in your own data.

AI Assistant Referral Traffic: where it shows up in GA4 and why it hides

AI assistant visits land in GA4 in one of three ways, depending on what the click carries with it. Each way sends the session to a different place in your reports, which is why the traffic is easy to miss.

Referrer, UTM-tagged and untagged visits

A visit from an assistant can arrive with a referrer, with UTM parameters, or with no source information at all. When the link passes a referrer, GA4 reads the referring domain and usually files the session as Referral. When the link carries UTM parameters, GA4 uses the source and medium in the tag, whatever the referrer says. When the app strips the referrer, or opens your page in its own in-app browser, GA4 has nothing to go on and the visit drops into Direct or Unassigned. Assistants change how they handle links, so don't take anyone's word for how a given one tags its clicks. Check your own reports.

Source and medium reports for AI assistant domains

Source and medium reports are the first place to find assistant visits. In GA4, go to Reports, then Acquisition, then Traffic acquisition, and set the dimension to Session source / medium.

Scan the list for domains that belong to AI assistants, such as chatgpt.com or gemini.google.com, usually paired with "referral". Write down the exact source strings you see, because those are what your filters will need. If you see a domain you don't recognize, look it up before you count it. A domain like bing.com could be an assistant or a plain search click, so I keep it out of the assistant bucket unless something else confirms it.

A custom channel group for AI assistants in GA4

A custom channel group in GA4 puts every assistant domain under one label, so you stop hunting through raw rows each month. Go to Admin, then Data display, then Channel groups, and create a new group.

Add a channel called "AI Assistants" with a rule that matches session source against the domains you found in the previous step. Place it above Referral in the order, otherwise GA4 will classify those sessions as Referral first. Save it, and the new channel appears in your acquisition reports next to Organic Search and Paid Search. When a new assistant domain shows up in your source list, add it to the rule. If you'd rather not touch the shared configuration, a saved comparison or segment in an exploration does the same job for your own analysis.

UTM parameters on assistant links: how to read them

UTM parameters on assistant links tell you what the link itself claimed about its origin. Open the Traffic acquisition report, switch the dimension to Session source / medium, and add Session campaign as a secondary dimension.

Read the source and the medium together. The source names who sent the visitor; the medium tells you how GA4 will classify the session. An assistant might add a source tag without a medium, or a medium your channel rules don't recognize, and then the session ends up in Unassigned. If you see tagged sessions sitting in Unassigned, add a matching condition to your AI Assistants channel. Don't assume any particular assistant always tags its links. Look at what actually arrives.

Landing pages as a second signal for assistant visits

Landing pages give you a second clue when the source is missing. In GA4, open Engagement, then Landing page, and compare it with your Direct and Unassigned traffic.

An assistant visit may follow a specific question, and its landing page may be the one that answers it: a service page, a location page or a pricing explanation. A direct visit can also come from someone who types your business name and lands on the home page. Take a Laval dental clinic as a hypothetical example. If its page about emergency appointments has direct visits with no campaign behind them, an assistant answer is one possible explanation. That isn't proof. Still, it tells you where to look, and our AI visibility check can show whether assistants are mentioning the business for those questions at all.

A related trap: someone reads an assistant's answer, then searches your business name on Google. That visit shows up as branded organic search, not as an assistant visit. Keep an eye on branded queries in Search Console alongside your assistant channel, and don't credit either one with the other's visits.

Conversion and call tracking for visitors from AI assistants

Conversion tracking turns assistant visits into something you can judge, because a visit only matters if it becomes a form submission, a call or a quote request. Mark those events as key events in GA4, then break them down by your AI Assistants channel.

Phone calls need extra work, since not every visitor fills in a form. Dynamic number insertion shows a different tracking number depending on how the visitor arrived, so a call from an assistant visitor gets recorded against that source instead of being guessed at. This is the kind of setup we handle in our call and conversion tracking work. Without it, you can count the visits but you can't tell whether any of them turned into business.

Loi 25 consent and the limits of the numbers

Loi 25 consent rules in Québec limit what your analytics can record. Your site needs a consent banner, and analytics only measures fully once a visitor accepts it.

Visitors who decline, or who close the banner without choosing, are partly or entirely missing from your reports, and assistant visitors are no exception. Add the sessions that lose their referrer, and the branded searches that follow an assistant answer, and the conclusion is simple: your assistant numbers are a partial view. Use them for trends and direction, not exact counts. I'm not giving legal advice here, so check your consent setup with someone qualified.

Monthly AI referral tracking checklist for a Québec business

A monthly AI referral tracking checklist keeps the picture honest without turning it into a full-time job. Once a month I'd run through these steps:

1. Review Session source / medium for new assistant domains and add any you confirm to the AI Assistants channel.
2. Check Unassigned for UTM-tagged sessions your rules missed.
3. Compare landing pages for Direct and Unassigned traffic against last month.
4. Look at branded queries in Search Console, kept separate from the assistant channel.
5. Break key events and tracked calls down by channel.
6. Note anything that changed on the consent banner, since it affects what gets recorded.
7. Log the month in one sheet so trends show over time.

If you want to understand how being found on Google and being cited by assistants fit together, our Found on Google, cited by AI system explains the approach.