A customer calls and says, “ChatGPT told me to call you.” You open Google Analytics to see how often that happens, and there’s nothing. No ChatGPT row, no AI anything. Just Referral, Direct and Organic Search, the same as last year.
The visits are probably in there. They’re filed under names you wouldn’t think to look for, and some of them are filed under nothing at all. In May 2026 Google added a built-in AI Assistant channel to GA4 (Google Analytics 4, the free version of Google’s site analytics). That helps, but it still misses some of what you’d want to count, and it can’t see anything from before it launched.
This post shows you how to track AI traffic in GA4 properly. You’ll get the list of AI sites to watch and a report that shows which pages they send people to. Then a pattern you can paste into GA4 so it keeps counting, and a realistic sense of what the numbers can and can’t prove.
Why your reports hide AI visitors
Every visit to your site carries a note about where it came from, as long as the visitor’s browser passes one along. That note is called the referrer. When someone clicks a link inside ChatGPT on a laptop, the referrer usually says chatgpt.com, and GA4 files the visit under that name.
Before May 2026, GA4 lumped those visits into Referral, right next to the chamber of commerce directory and the supplier who links to you. Unless you went digging, you’d never see them as their own line.
Three things still hide them today. Google’s new AI Assistant channel only covers the assistants Google recognizes, and Google hasn’t published that list beyond naming ChatGPT, Gemini and Claude. Visits from phone apps often arrive with no referrer at all, so GA4 calls them Direct, the same bucket as someone typing your web address. And the new channel doesn’t rewrite history, so everything before mid-May 2026 is still sitting in Referral.
If you’re still working on getting mentioned by these tools in the first place, our piece on how businesses get cited by ChatGPT and Perplexity covers that side. This post is about counting the people who show up afterward.
The AI sites to look for in your traffic
Before you build anything, check what’s already there. In GA4, go to Reports, then Acquisition, then Traffic acquisition. Change the table’s first column from the channel to Session source, then search the table for these names:
- ChatGPT shows up as chatgpt.com, or chat.openai.com on older links.
- Perplexity shows up as perplexity.ai.
- Google Gemini shows up as gemini.google.com.
- Claude shows up as claude.ai.
- Microsoft Copilot shows up as copilot.microsoft.com or copilot.com.
- DeepSeek shows up as chat.deepseek.com.
- Grok shows up as grok.com, and Meta AI as meta.ai.
ChatGPT is the easiest one to follow. OpenAI adds a small tag to the end of the links it shows people (utm_source=chatgpt.com), so GA4 can name the source even when the browser drops the referrer.
These names change more often than you’d think. Bing Chat became Copilot, Bard became Gemini, and chat.openai.com moved to chatgpt.com. Plan on checking this list a couple of times a year.
If half those names are new to you, that’s worth fixing too, because your customers are using them to shortlist businesses like yours. We also train business teams to use AI tools in their everyday work, which is the fastest way to see what a customer sees when they ask about you.
Build a report that shows what AI visitors do
The standard reports tell you how many people came. An exploration, which is GA4’s build-your-own report, tells you which pages they landed on and whether they went on to call or fill out a form. That second part is what tells you if AI traffic is worth anything to your business.
- Click Explore in the left menu and start a Blank exploration.
- Set the date range to the last 90 days. AI traffic is thin for most small businesses, and a single week will mislead you.
- Next to Dimensions, click the plus sign. Import Session source and Landing page + query string.
- Next to Metrics, click the plus sign. Import Sessions, Engaged sessions and Key events.
- Drag Session source into Rows, then drag Landing page + query string in underneath it.
- Drag Sessions, Engaged sessions and Key events into Values.
- Under Filters, choose Session source, pick “matches regex,” and paste the pattern from the next section.
Key events, if you haven’t met the term, are whatever GA4 counts as a win on your site, like a sent form or a tapped phone number. An engaged session is a visit where someone stuck around or clicked further, as opposed to landing and leaving.
You now have a table of every AI source, the pages each one sent people to, and what those people did next. Name it something you’ll recognize next month, like “AI visitors by page.”
Look at the landing pages first. When an assistant recommends you, it usually links to one specific page, and it often isn’t your home page. If the same service page keeps showing up, that page is doing something the assistants like. Keep it accurate and keep it current, because it’s now one of your front doors.

The pattern to paste into GA4
A regex, short for regular expression, is a search pattern. Instead of setting up eight separate filters, you give GA4 one line that means “any of these sites.” GA4 treats it as an all-or-nothing match on the whole source name, which is why this one has a start marker (^) and an end marker ($).
^(.*.)?(chatgpt.com|chat.openai.com|perplexity.ai|claude.ai|gemini.google.com|copilot.microsoft.com|copilot.com|deepseek.com|grok.com|meta.ai|chat.mistral.ai|you.com|poe.com)$ A few notes before you paste it. The backslash in front of each dot tells GA4 to read it as a plain dot and not a wildcard. The (.*.)? at the front lets it catch versions with something added in front, like www. or chat., so chat.deepseek.com still counts. And because it has to match the whole name, a site like bayou.com won’t sneak in under you.com.
Google’s own help page offers a much looser example that catches anything ending in “ai” or containing “gpt.” That will sweep in sites that have nothing to do with AI assistants, so we’d skip it and use a list of real names.
Save this line in a note somewhere you’ll find it. A filter you build in one exploration doesn’t carry over to the next one, and you’ll want this pattern again for the step below.
Set up a channel so it keeps counting on its own
The exploration answers the question once. A custom channel group makes AI visitors show up in your everyday reports without rebuilding anything. It also reaches back before May 2026, because Google applies custom channel groups to past data in your reports.
- Go to Admin, then Data display, then Channel groups.
- Click Create new channel group. GA4 starts you with a copy of the default channels, which is what you want.
- Give it a plain name, like “Channels with AI.”
- Click Add new channel and call it “AI assistants.”
- Add a condition: Source, matches regex, and paste the pattern above.
- Add a second condition joined with OR: Medium, exactly matches, ai-assistant. That picks up any assistant Google adds to its own list later.
- Save the channel. Click Reorder and drag “AI assistants” above Google’s built-in AI Assistant channel and above Referral.
- Save the group.
The reorder step is the one people miss. GA4 hands each visit to the first channel on the list that matches, working from the top down. If Referral sits above your new channel, Referral keeps every AI visit, your channel reads zero, and you’ll assume the pattern is broken.
To see the result, go back to Traffic acquisition and switch the table’s first column to your new channel group. A standard GA4 property only gets two custom channel groups, so if you already use one, add the AI channel to it instead of spending your last slot.
What small numbers can and can’t tell you
For a lot of local businesses, AI visits are still a small slice of traffic, sometimes a handful a month. That’s normal. It’s no reason to rebuild your marketing around them, and it’s no reason to ignore them either.
Small numbers swing hard. Going from three visits to six looks like a doubling on a chart and means almost nothing. Compare 90-day stretches instead of weeks, and don’t trust a percentage until there are a few dozen visits behind it.
The fair objection is that this is a lot of setup for a few visits. If AI sends you eight people a month, why bother? We’d give two reasons. Those visitors arrive after an assistant has already described you and pointed them at a specific page, so they can be further along than someone scrolling search results. And the setup takes about an hour once, while the history it builds is something you can’t go back and collect later.
Count leads, not visits. Five AI visitors who fill out your quote form are worth more than fifty who leave. If your forms and phone taps aren’t tracked as key events yet, start with our guide to tracking forms, calls and booked consultations in GA4 without double counting, then come back to this.
What GA4 still can’t see
Even with all of this in place, GA4 only counts people who click. A lot of AI use never produces a click. Someone asks Perplexity for a good roofer in Cedar Falls, reads your name in the answer, and calls you straight from their phone. GA4 never hears about it.
Clicks from Google’s own AI answers, the AI Overviews at the top of search results and the newer AI Mode, show up as regular Google organic search. GA4 doesn’t split them out, and your channel group can’t either.
App visits that arrive without a referrer land in Direct, and no pattern can pull them back out. GA4 also can’t tell you what question the person asked the assistant, which is often the most useful thing to know.
The cheapest fix for most of this is old-fashioned. Add “How did you hear about us?” to your contact form and include “ChatGPT or another AI assistant” as one of the answers. Ask on the phone, too. Between those answers and your new channel, you’ll have a much more honest picture than either one gives you alone.

Where to start this week
If you only do one thing, open Traffic acquisition, switch to Session source, and look for the names on the list above. That takes five minutes and tells you whether there’s anything to track. If there is, build the channel group so it keeps counting while you get on with running the business.
GA4 isn’t the only way to see this. If you’re a Wapiti client, you can also use Wapiti Analytics, our own privacy-respecting alternative to Google Analytics. It puts goal conversions, UTM campaign tracking and each visitor’s path through your site on one dashboard, for owners who want the answers without learning GA4’s menus. And if you’d like a hand with the GA4 setup in this post, ask us and we’ll check that your forms and calls are counted right too.

