AI assistants are quietly sending visitors to websites, and most of that traffic is sitting in your Google Analytics 4 property right now, unlabeled, lumped into the generic Referral bucket. This guide shows you exactly where to find it, how to give it its own channel so it shows up in your standard reports, and why the number you see is a floor, not the whole picture.
What counts as AI referral traffic?
AI referral traffic is any session that starts when someone clicks a link inside an AI answer engine — ChatGPT, Perplexity, Microsoft Copilot, Gemini, Claude — and lands on your site. When these tools cite a source, they usually link to it, and when a user clicks that link in a web browser, the assistant's domain is passed along as the referrer. GA4 records it like any other referral: source = the AI platform's domain, medium = referral.
That's the good news. The bad news is that GA4 has no built-in "AI" channel, so this traffic gets mixed in with every other referral — partner sites, directories, that one forum thread from 2019. If you want to see the trend, you have to separate it yourself.
Which referral sources should you look for in GA4?
Open Reports → Acquisition → Traffic acquisition, change the primary dimension to Session source / medium, and search for these domains:
- chatgpt.com — ChatGPT's current domain. Older sessions may show chat.openai.com, the legacy domain, so check both.
- perplexity.ai — Perplexity cites sources prominently, so it often over-indexes as a referrer relative to its user base.
- copilot.microsoft.com — Microsoft Copilot's web interface.
- gemini.google.com — Google's Gemini assistant. Note this is separate from clicks on Google AI Overviews, which arrive as ordinary google / organic and cannot be isolated in GA4.
- claude.ai — Anthropic's Claude, relevant if your audience skews technical or professional.
If any of these appear, you're already receiving AI referrals. Even a handful of sessions per week is a signal worth tracking, because the mix of platforms tells you where you're being cited — which pairs nicely with the manual approach in our guide to tracking AI citations.
How do you build a custom channel group for AI referrals?
The cleanest solution is a custom channel group, which makes "AI Referrals" a first-class channel in your standard reports. In GA4:
- Go to Admin → Data display → Channel groups.
- Click Create new channel group (GA4 copies the default group as a starting point — don't edit the default itself).
- Add a new channel named AI Referrals and drag it to the top of the list, because GA4 assigns traffic to the first channel whose conditions match.
- Set the condition: Source matches regex chatgpt\.com|chat\.openai\.com|perplexity\.ai|copilot\.microsoft\.com|gemini\.google\.com|claude\.ai
- Save. From that point on, any report that supports channel grouping can display AI Referrals alongside Organic Search, Direct, and the rest.
One caveat: custom channel groups apply from creation onward in some views and retroactively in others, so build yours now even if the volume is tiny. Future-you will want the history.
How do you build an exploration for deeper analysis?
For ad-hoc analysis, use a Free-form exploration. Go to Explore, create a new exploration, and set it up like this: add Session source as a dimension and Sessions, Engaged sessions, Average engagement time per session, and your Key events as metrics. Then add a filter — Session source matches the same regex from the channel group. Add Landing page as a second dimension to see which pages the AI engines are actually citing.
The landing-page view is the most actionable report in this whole exercise. If Perplexity keeps sending people to one specific FAQ page, that page's structure is working — study it and replicate it. That's the feedback loop that makes Answer Engine Optimization improvable rather than guesswork.
Why is AI referral traffic small but disproportionately valuable?
Expect single-digit percentages of your total traffic, often less. That's normal, and it's not the right way to judge this channel. Two reasons:
- Pre-qualified visitors. Someone who clicks through from an AI answer has already read a summary of what you do and decided they want more. They arrive mid-funnel or later. Many site owners report that AI referral sessions engage longer and convert at noticeably higher rates than average — check your own engagement and key-event columns rather than taking anyone's word for it.
- The click is the tip of the iceberg. Most people who see your brand in an AI answer never click at all — the same dynamic behind zero-click searches. The referral sessions you can measure are a sample of a much larger volume of AI-driven brand exposure.
What AI traffic will GA4 never show you?
A large share of AI-influenced visits arrives as "dark traffic" — sessions with no usable referrer. Know the gaps so you don't misread your data:
- Native apps. Clicks from the ChatGPT mobile or desktop apps often pass no referrer and land in Direct.
- Voice interactions. If an assistant answers out loud, there's no click to measure. The user may later type your brand into Google — which shows up as branded organic search, not AI.
- Copy-paste behavior. Users frequently copy a URL or brand name from an AI answer and paste it into the address bar. That's Direct traffic in GA4, invisible as AI-influenced.
- Google AI Overviews. Clicks from AI Overviews are reported as regular google / organic. GA4 gives you no dimension to split them out.
A practical proxy: watch Direct traffic and branded search volume alongside your AI Referrals channel. If all three rise together after you start showing up in AI answers, the referral number is confirming a bigger trend, not capturing all of it.
How do you set a baseline before starting AEO work?
Before you change anything on your site, record where you stand — otherwise you'll never be able to prove the work paid off. A minimal baseline takes about thirty minutes:
- Build the channel group and exploration described above.
- Export the last 90 days of AI referral sessions, engaged sessions, and key events, broken down by source and landing page.
- Note your current Direct traffic and branded-search levels as context for the dark-traffic effect.
- Record the date you begin AEO changes (GA4 removed manual annotations, so keep a simple changelog in a spreadsheet).
Then measure the same numbers monthly. AI referral growth tends to be lumpy — a single new citation on a high-traffic answer can double your monthly sessions from one platform — so judge trends over quarters, not weeks.
Frequently asked questions
Why don't I see chatgpt.com in my GA4 reports at all?
Either AI engines aren't citing you yet, or they're citing you without links being clicked. Check whether your content is accessible to AI crawlers and structured so it can be quoted — our DIY AEO audit checklist covers the most common blockers.
Can I track Google AI Overviews traffic separately in GA4?
No. Clicks from AI Overviews arrive as standard google / organic sessions, and GA4 offers no dimension to separate them from classic search clicks. Google Search Console blends AI Overview impressions into regular search data as well.
Is AI referral traffic in GA4 accurate?
It's directionally reliable but systematically undercounted. App-based clicks, voice answers, and copy-paste visits all land in Direct or branded search. Treat the AI Referrals channel as a minimum bound on your AI-driven traffic.
How much AI referral traffic is a good result?
There's no universal benchmark — it varies wildly by industry and by how much of your audience uses AI tools. What matters is the trend against your own baseline and the conversion rate relative to your other channels. If you'd like an outside read on why the number is low, our free AEO audit checks whether AI engines can crawl, parse, and cite your site — and shows you exactly what's blocking them.