How to track AI-referred traffic in GA4 (ChatGPT, Perplexity, and others)

Three months ago I noticed something in a client’s GA4 data that made me do a double-take. Their organic search traffic was up 12% year over year, but Google Search Console showed flat impressions and clicks. The traffic was real. It was converting. But it wasn’t coming from Google Search.

It was coming from ChatGPT.

When I dug into the referral data, I found that chat.openai.com was their fifth-largest referral source. It had been growing steadily for months. Nobody on their team had noticed because GA4 was classifying it as regular referral traffic, buried in a long list of domains that nobody reviewed.

AI-generated referrals are a real and growing traffic source. If you’re not tracking them separately, you’re missing a category of traffic that behaves differently from everything else in your analytics.

Where AI traffic actually comes from

Let’s start with which AI tools send web traffic and how they do it.

ChatGPT (chat.openai.com): When ChatGPT cites sources or links to pages in its responses, users click those links. The referrer is chat.openai.com. This is now one of the top AI referral sources for most sites I look at. Since OpenAI added browsing capabilities and citations became standard, the click-through volume has been growing month over month.

Perplexity (perplexity.ai): Perplexity is built specifically as an answer engine with citations. Every response includes numbered source links. Users click them frequently. Referrer is perplexity.ai or www.perplexity.ai.

Google Gemini (gemini.google.com): Google’s AI chat product links to sources. Interesting wrinkle here: some Gemini-referred traffic shows up with a gemini.google.com referrer, but traffic from AI Overviews in Google Search gets mixed into regular Google organic traffic. There’s no easy way to separate AI Overview clicks from standard search clicks in GA4. Google hasn’t provided that distinction.

Claude (claude.ai): Anthropic’s Claude occasionally cites sources when browsing. Referrer shows as claude.ai. The volume is lower than ChatGPT and Perplexity in my data, but it’s growing.

Microsoft Copilot (copilot.microsoft.com): Copilot integrates with Bing search results and provides linked sources. Referrer is copilot.microsoft.com.

Other sources: You’ll also see traffic from you.com, phind.com, and various other AI search tools. The long tail is real and growing.

How GA4 classifies AI traffic (badly)

Here’s the problem. GA4’s default channel groupings don’t have an “AI” channel. There’s Organic Search, Paid Search, Direct, Referral, Social, Email, and a few others. AI-referred traffic gets dumped into one of three buckets:

Referral. Traffic from chat.openai.com, perplexity.ai, and claude.ai typically lands in the Referral channel. This is technically correct but unhelpful. Referral is a catch-all for any domain that isn’t specifically classified as search or social. Your AI traffic is mixed in with random blog mentions, forum links, and partner sites.

Organic Search. Some AI-referred traffic, especially from Google’s own Gemini AI Overviews, gets classified as Organic Search. This inflates your organic search numbers and makes it look like SEO is driving growth when it’s actually AI referrals.

Direct. When AI tools don’t pass a referrer (some mobile app versions don’t), the traffic falls into Direct. This is the analytics junk drawer, and I’ve written about the problems with GA4’s default classifications before.

None of these classifications help you understand AI as a distinct traffic category. You need a custom setup.

Identifying AI referral domains

First, let’s find out what AI traffic you’re already getting. Go to GA4 > Reports > Acquisition > Traffic Acquisition. Add a secondary dimension of “Session source.” Sort by sessions descending and look for these domains:

• chat.openai.com
• chatgpt.com
• perplexity.ai
• www.perplexity.ai
• gemini.google.com
• claude.ai
• copilot.microsoft.com
• you.com
• phind.com
• poe.com
• bard.google.com (older Gemini referrer)

You might not find all of them. But I’d bet money you find at least two or three. Note the session counts. For most sites I audit in 2026, AI referrals account for 2-8% of total traffic. For content-heavy sites with strong topical authority, I’ve seen it as high as 15%.

If your traffic acquisition report is long and you’re having trouble finding these domains, use the search/filter at the top. Type “openai” or “perplexity” and see what shows up.

Setting up a custom AI channel grouping

GA4 lets you create custom channel groups that override the default classifications. Here’s how to create one that properly identifies AI traffic.

Step 1: Create the custom channel group

Go to Admin > Data display > Channel groups > Create new channel group.

Name it something descriptive. I use “Custom Channels (with AI)” so I can distinguish it from the default group in reports.

Step 2: Copy the default channels

GA4 will pre-populate the default channels. Keep all of them. You’re adding a new channel, not replacing the existing ones.

Step 3: Add an AI channel

Click “Add new channel” and name it “AI Referral” (or “AI” or “AI Traffic” depending on your preference for brevity).

Set the conditions:

• Source matches regex: chat\.openai\.com|chatgpt\.com|perplexity\.ai|gemini\.google\.com|claude\.ai|copilot\.microsoft\.com|you\.com|phind\.com|poe\.com|bard\.google\.com

That regex covers the major AI referral sources. You’ll want to update this list as new AI tools emerge. I revisit mine quarterly.

Step 4: Set channel priority

Drag the AI Referral channel above Referral and Organic Search in the priority list. Channel rules are evaluated in order. You want GA4 to check the AI conditions before falling back to the generic Referral or Organic Search classifications. If you don’t set the priority correctly, the traffic will continue to match the broader channels first and never reach your AI channel.

Step 5: Apply the custom channel group to reports

In your standard reports, you can switch from the default channel grouping to your custom one using the dropdown in the report header. In explorations, select your custom channel group as the dimension instead of “Default channel group.”

One limitation: custom channel groups in GA4 are not retroactive for the channel grouping dimension itself. They apply from the moment you create them going forward. However, the underlying source/medium data is historical, so you can use segments and filters on source to analyze historical AI traffic even before you created the custom channel.

Creating an AI traffic segment and audience

Beyond channel groupings, you’ll want a segment for explorations and an audience for Google Ads.

Segment

In any Exploration, create a new session segment:

• Condition: Session source matches regex chat\.openai\.com|chatgpt\.com|perplexity\.ai|gemini\.google\.com|claude\.ai|copilot\.microsoft\.com|you\.com|phind\.com|poe\.com

Name it “AI Referral Sessions.” Now you can apply this segment to any exploration and compare AI traffic behavior to other sources.

Audience

Go to Admin > Audiences > New audience > Create a custom audience.

Same condition as the segment. Save it as “AI Traffic” and enable the Google Ads audience sharing toggle if you want to use it for targeting or exclusions. As I covered in my piece on GA4 audiences most people ignore, building audiences for emerging traffic sources is one of the highest-ROI things you can do with your analytics setup.

A useful secondary audience: “AI Traffic Converters.” Same AI source condition, but add a condition that the user completed a conversion event. This tells you not just how much AI traffic you’re getting, but how much of it is actually valuable.

What AI traffic means for your content strategy

I want to share some patterns I’m seeing across client properties, because AI traffic doesn’t behave like other traffic.

Engagement is different. AI-referred users tend to have specific intent. They asked a question, got an answer with a citation, and clicked to verify or learn more. Average engagement time for AI referral sessions is typically 20-40% higher than organic search in my data. These aren’t casual browsers. They’re people who want depth.

Bounce patterns are different. The “bounce” concept is less useful here. Many AI-referred visitors read the specific page they landed on thoroughly and then leave. That looks like a bounce but it’s actually a satisfied user. Look at engagement time and scroll depth instead of bounces.

Conversion behavior varies by tool. In my data, Perplexity users convert at the highest rate among AI referrals. My theory: Perplexity users are in active research mode and ready to take action. ChatGPT users are more exploratory. But this varies significantly by industry and content type.

Content type matters. AI tools disproportionately cite content that’s structured, factual, and directly answers specific questions. Long-form guides, data-backed articles, and pages with clear definitions rank well in AI citations. Opinion pieces and branding content don’t get cited as much.

What this means for your content strategy: if AI referrals are a meaningful traffic source (say, above 3-5% of total traffic), consider optimizing for it. That means:

• Structuring content with clear headings and direct answers
• Including data points and statistics that AI tools can cite
• Maintaining factual accuracy (AI tools have started penalizing sites with inconsistencies)
• Building topical authority around specific subjects rather than covering everything superficially

My data: what I’m seeing across client properties

I’ll share some aggregate numbers from properties I manage, without revealing specific clients.

Traffic share: Across 20+ properties in different industries, AI referrals account for an average of 4.7% of total sessions in Q1 2026, up from 1.2% in Q1 2025. Growth has been roughly 3-4x year over year.

Top sources by volume:

1. ChatGPT (chat.openai.com + chatgpt.com): 45% of AI referrals
2. Perplexity: 28% of AI referrals
3. Google Gemini: 12% of AI referrals
4. Copilot: 8% of AI referrals
5. All others: 7%

Geographic patterns: AI referral traffic is disproportionately US-based across my clients. About 60% of AI referral sessions come from the US, compared to 35% for overall traffic on the same properties. This is shifting as AI tools become more popular globally.

Device split: 55% mobile, 45% desktop. AI referral traffic skews more mobile than overall traffic, which tracks with the growing usage of ChatGPT and Perplexity mobile apps.

Conversion rates: AI referral traffic converts at roughly 1.5x the rate of standard referral traffic and 0.8x the rate of organic search. So it’s better than most referral sources but slightly below organic search. Keep in mind that organic search benefits from intent matching (the user searched for something specific), while AI referrals are more variable. The important thing is that UTM tagging doesn’t apply here since you don’t control the links AI tools generate, which makes proper referral detection even more important.

Month-over-month trend: Growing consistently. I haven’t seen a single property where AI referral traffic declined quarter over quarter. The trajectory is clearly upward.

What to do right now

If you’ve read this far, here’s your action list:

1. Check your GA4 right now. Look at Traffic Acquisition > Session source and search for the AI domains listed above. See what you’re already getting.

2. Create the custom channel grouping. Takes 15 minutes. You’ll immediately get cleaner reporting.

3. Build the audience. Even if you’re not sure what to do with it yet, start collecting the data. Audiences aren’t retroactive.

4. Review monthly. Add new AI referral domains as they emerge. The list I gave you covers 2026 well, but this space moves fast. Check your referral report for unfamiliar domains regularly.

5. Don’t panic about SEO. AI traffic is largely additive, not cannibalistic. In most cases I’ve seen, it’s new traffic from users who would have asked a friend or searched Reddit, not traffic that would have come through Google Search. Think of it as a new channel opening up, not an existing channel being redirected.

AI tools are becoming how people discover content. Your analytics should reflect that reality.