How to track ChatGPT and AI search traffic in GA4

To track ChatGPT traffic in GA4, create a custom channel group with an "AI assistants" channel that matches sources like chatgpt.com, perplexity.ai, claude.ai, gemini.google.com and copilot.microsoft.com, then move it above Referral. It takes about 15 minutes and applies to past data. It only counts clicks, though. Mentions without a click and Google's AI Overviews need other measures.

Why your AI traffic is hiding in Referral

A founder asks their team a reasonable question: "Are we getting anything from ChatGPT?" Someone opens GA4, looks at the channel report, and finds Organic Search, Direct, Paid, Email and Referral. No AI. So the answer becomes a shrug, and the shrug becomes a budget decision.

The traffic is usually there. GA4's default channel group has no AI bucket, so a visit from chatgpt.com gets filed under Referral next to a link from a partner blog. It's like counting birds at a feeder with one tally for "everything with feathers."

Think of GA4 as a trail camera on a game trail. It's very good at recording what walks past the lens. It records nothing about what walked around it. Setting up an AI channel points the camera at the right trail. Then we'll talk about what goes around it, because that's where most of AI's influence lives. (The tracks in the mud, not the photos.)

Step 1: check what sources you're already getting

Before you build anything, look at the actual source names in your property. It takes two minutes and saves you from writing a rule for domains you never see.

1. In GA4, open Reports → Acquisition → Traffic acquisition.
2. Change the primary dimension to Session source.
3. Search the table for "chatgpt", "perplexity", "claude", "gemini" and "copilot".

ChatGPT makes this easier than most. OpenAI says ChatGPT adds utm_source=chatgpt.com to referral links from its search results (OpenAI), so those visits show up with a clean source. Write down every AI-looking source you find, including odd variants.

Step 2: build the AI assistants channel

GA4 lets a standard property create two custom channel groups, each with up to 50 channels (Google Analytics Help). You'll use one.

1. Go to Admin → Data display → Channel groups.
2. Find the Default channel group and choose Copy to create new. Starting from a copy keeps every other channel working the same way.
3. Name it something plain, like "Channels with AI".
4. Click Add new channel and name it AI assistants.
5. Set the condition to Session source, matches regex, and paste:

^(.*\.)?(chatgpt\.com|chat\.openai\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|copilot\.microsoft\.com)$

1. Add any extra sources you found in Step 1 to the list inside the parentheses, with dots escaped the same way.
2. Save the channel, then use Reorder to drag AI assistants above Referral.
3. Save the group.

That reorder step is the one people skip. Google is blunt about it: "Traffic is included in the first channel whose definition it matches given the current order of channels in the group" (Google Analytics Help). Leave AI below Referral and Referral eats it first. Your new channel sits there empty and you conclude nobody uses ChatGPT.

The anchors (^ and $) keep the rule tight, so a source only matches if it's genuinely one of those domains or a subdomain of one.

Step 3: read it in your reports

Google says custom channel groups "can be applied to your reports retroactively" (Google Analytics Help), so you don't have to wait a month for data. In the Traffic acquisition report, switch the primary dimension to your new channel group and set the date range back as far as you like.

Then look past the session count. Three questions tell you more than volume:

• Which pages do AI visitors land on? That's what the engines are citing. It's often a comparison page, a pricing page or one deep how-to article, not your homepage.
• Do they convert? Compare demo requests or sign-ups per session against organic search. AI visitors may arrive further along, since the chat already did some of the shortlisting. Your own numbers will tell you if that's true for you.
• Which engine sends the most? If one engine sends nearly all of it, check your visibility in the others before you assume they don't matter.

What GA4 can't see: three blind spots

Now the honest part. A trail camera sees the trail it's pointed at. AI influence mostly walks around it, in three ways.

Blind spot 1 is the big one. AI answers often name products without the buyer clicking anything. They open a new tab and type your name a day later. GA4 calls that Direct or branded Organic. So your AI channel will always undercount AI's real effect. The way to see the tracks is to check the answers themselves: run the same 10–20 buyer prompts each month in ChatGPT, Perplexity and Gemini and log each time you're named. Our guide on why ChatGPT recommends your competitor has a simple way to score what you find.

Blind spot 2 trips up nearly everyone. Google says sites in AI Overviews and AI Mode "are included in the overall search traffic in Search Console," reported "within the 'Web' search type" (Google Search Central). There's no separate line for it. Your regex will never catch it, because the source is just Google.

Blind spot 3 is the oldest problem in analytics. When a click arrives without referrer information, GA4 has nothing to file it under except Direct. A free-text "How did you hear about us?" field on your demo form is crude, and it's still the best tool here. Buyers will happily type "ChatGPT told me" when no tag would ever record it.

The three gauges: a measurement setup you can defend

Put those together and you get three gauges, each covering what the others miss:

1. Clicks: the AI assistants channel in GA4. Exact, and always an undercount.
2. Mentions: a monthly prompt check across the main engines. Tells you if you're in the answer at all.
3. Echoes: branded search in Search Console, Direct traffic trends and self-reported attribution on forms. Shows the influence that never clicked.

When all three rise together, your AEO work is working. When clicks are flat but mentions and echoes climb, it's still working, and the trail camera just isn't pointed at it.

If you haven't started on the visibility side yet, begin with what AEO means for B2B software, then the full method in how to get your SaaS recommended by ChatGPT. And if you'd like help setting up the gauges, it's part of how we run analytics and reporting for clients.