How to Track AI Search Traffic in GA4 (ChatGPT, Perplexity & Gemini)

AI assistants now send real visitors to your site — people who asked ChatGPT, Perplexity or Gemini a question, clicked a cited link, and landed on your page ready to act. The problem is that most of that traffic hides inside your "Referral" and "Direct" buckets in Google Analytics 4, so it looks like nothing is happening. If you cannot see it, you cannot prove its value or optimise for it.

This guide walks through how to track AI search traffic in GA4 the way it actually works in 2026: the new native AI Assistant channel Google added in May, a custom channel group that catches the platforms Google's default rules miss, and the built-in undercount you need to disclose to anyone reading the numbers. No tag manager wizardry required — most of this is a five-minute settings change.

Why AI traffic is invisible by default

When someone clicks a link inside an AI product, their browser usually passes a referrer — the domain the click came from, like chatgpt.com or perplexity.ai. GA4 reads that referrer and files the session under a channel. Historically, none of those AI domains matched GA4's built-in rules, so the visits fell through to Referral, mixed in with newsletters, forums and every other link on the open web.

That is why AI traffic feels like a rumour in most analytics accounts. It is genuinely there — a 2026 Previsible study of 6.77 million sessions found standalone chatbot referrals growing nearly 10x — but without a rule that isolates it, the number is buried. Our breakdown of that study, ChatGPT's 92% share of AI referral traffic, explains just how much volume is at stake. The fix is to teach GA4 which domains count as AI.

Option 1: Use GA4's native AI Assistant channel

On May 13, 2026, Google added a native AI Assistant channel to GA4's Default Channel Group. When an incoming session's referrer matches a recognised AI domain, GA4 now tags it with the medium ai-assistant and files it under this channel automatically. There is nothing to switch on — it is already collecting data in every property.

To see it:

1. Open Reports → Acquisition → Traffic acquisition.
2. Set the primary dimension to Session default channel group.
3. Look for the AI Assistant row.

It is the fastest way to get a directional read, and because it is retroactive on the medium, you can often see historic sessions too. But there is a catch that makes it insufficient on its own.

The limitation: it only knows three tools

Google's published list for the AI Assistant channel names just ChatGPT, Gemini and Claude. Visits from Perplexity, Copilot, Meta AI and others are not captured — they stay stranded in Referral. Given that Perplexity and Copilot still send meaningful traffic to some verticals, relying only on the native channel means undercounting the AI channel and misattributing the rest. For a complete picture, build a custom channel group as well.

Option 2: Build a custom AI channel with regex

A custom channel group lets you define exactly which referrers count as AI and update the list as new platforms appear. This is the durable, comprehensive method.

1. Go to Admin → Data settings → Channel groups (property level).
2. Click Create new channel group and name it clearly — AI Traffic (2026) works, since you will revise the regex over time.
3. Add a new channel called AI Search.
4. Set the condition to Source matches regex, and paste the pattern below.

chatgpt\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|copilot\.microsoft\.com|chat\.openai\.com|bard\.google\.com|meta\.ai

That pattern captures the eight most significant AI referral sources currently passing attribution data to GA4. The escaped dots (\.) matter — an unescaped dot in regex means "any character," which would create sloppy matches.

The one step people forget

Drag the new AI Search channel above the Referral channel. GA4 evaluates channel rules in order, top to bottom, and stops at the first match. If Referral sits higher in the list, it fires first and your AI sessions never reach your custom rule — they stay misclassified. Order is not cosmetic here; it decides whether the whole setup works.

Once saved, the custom channel group becomes available as a dimension in Explorations and in the report dimension picker (it is not retroactive in standard reports the way the native medium is, so give it a few days of forward data before drawing conclusions).

Native channel vs. custom group: which to use

You do not have to choose — most teams run both. But it helps to know what each does well.

Factor	Native AI Assistant channel	Custom regex channel group
Setup effort	None — already on	~5 minutes
Platforms covered	ChatGPT, Gemini, Claude only	Every domain you add (8+)
Retroactive	Yes (via the medium)	No — forward-looking
You control the list	No — Google decides	Yes — edit any time
Best for	Quick directional check	Reporting and optimisation

Use the native channel for a fast gut-check today, and the custom group as your reporting source of truth going forward. When a new assistant gains share, you add one term to the regex and your reporting keeps pace — no waiting for Google to update its list.

The undercount you must disclose

Here is the uncomfortable truth every AI-tracking method shares: you will always capture less AI traffic than actually occurred. Across platforms, an estimated 35–70% of AI referral sessions arrive with no referrer header at all — often because the click came from a native mobile app, a stripped privacy context, or a copy-pasted link. Those sessions land in Direct no matter how good your channel rules are.

What this means in practice:

• Treat your AI channel number as a floor, not a total. Real AI influence is higher — sometimes much higher.
• Watch the trend, not the absolute. A rising AI channel alongside rising Direct traffic to informational pages is a strong signal AI is sending more than your rules can see.
• Correlate with landing pages. If Direct sessions to deep, specific URLs (not your homepage) are climbing, that is classic unattributed AI behaviour — people rarely type a long article URL by hand.

And remember the single biggest source that no GA4 rule will ever isolate: Google AI Overviews. Those clicks are attributed to ordinary Google organic traffic, so they never appear in any AI channel. If you want the fuller Google-side picture, pair this setup with the Search Console AI performance report, which surfaces impressions inside AI Overviews and AI Mode that GA4 cannot.

What to do once you can see the traffic

Measurement is only useful if it changes what you do. Once your AI channel is live, three moves pay off quickly:

1. Segment AI sessions by landing page

Build an Exploration with your AI channel as a filter and Landing page as the dimension. This tells you which pages AI tools actually cite — and those are the pages worth deepening, updating and interlinking first.

2. Compare engagement, not just volume

AI-referred visitors often arrive mid-task and behave differently from organic search visitors. Compare engagement rate, conversions and pages per session across channels. If AI traffic converts well, that is your case for investing in AI search optimization; if it bounces, your landing experience needs work before you chase more of it.

3. Set an annotation and revisit the regex quarterly

The AI landscape shifts fast. Add a calendar reminder to review your regex every quarter, adding new assistants as they gain referral share and pruning any that go quiet. A tracking rule is only as current as the list behind it.

The takeaway

AI search traffic is real, growing, and — until you tell GA4 how to spot it — nearly invisible. Turn on the native AI Assistant channel for an instant read, build a custom regex channel group so you catch Perplexity, Copilot and whatever launches next, and place it above Referral so the rules fire in the right order. Then report honestly: the number you see is a floor, Direct hides a large chunk, and AI Overviews sit outside GA4 entirely. Do that, and "AI traffic" stops being a rumour and becomes a channel you can actually grow.