How to Track ChatGPT and Perplexity Traffic in GA4

Clients increasingly ask how much traffic they get from ChatGPT. It is a fair question and, out of the box, GA4 answers it badly. AI visits are scattered across Referral, Direct and sometimes Organic, mixed in with everything else. A few minutes of setup fixes most of that.

This guide covers what GA4 records for AI assistant visits, how to set up a reliable AI channel, why some AI traffic will always look like Direct, and how to present the numbers to clients without overclaiming.

How AI assistant visits arrive in GA4

When someone clicks a link inside ChatGPT, Perplexity, Claude, Gemini or Copilot, one of three things happens.

1. The referrer is passed. GA4 records the source as the assistant's domain, such as perplexity.ai, and the medium as referral.
2. A UTM parameter is added. ChatGPT commonly appends utm_source=chatgpt.com to links it cites. GA4 uses that as the source.
3. Nothing is passed. In mobile apps, desktop apps and some in-app browsers, the referrer is often stripped. GA4 records the visit as Direct.

So without any setup, AI visits are split between Referral and Direct, and nothing labels them as AI.

What about Google's AI features?

Clicks from Google AI Overviews and AI Mode come from Google and are counted as organic search. They will not appear in an AI assistant channel. For Google's AI features, use Search Console's performance and generative AI reports. See Google AI Mode SEO.

Has GA4 added an AI channel?

Several sources reported that GA4 began rolling out a built-in "AI Assistant" default channel during 2026, assigning recognised AI sources to it automatically. If the client's property has it, use it. Two caveats apply:

• It is not retroactive. Sessions recorded before the channel existed stay in the channel they were originally assigned.
• Coverage may lag. New assistants and new domains appear faster than any default list is updated.

Check the official GA4 documentation for current behaviour. Either way, a custom channel group gives you control and consistency across all client properties.

Set up a custom AI channel group

Custom channel groups let you define your own rules for classifying traffic and apply them to historical data in reports. Here is the setup.

Step 1: Create the channel group

In GA4, go to Admin → Data display → Channel groups and create a new channel group. Start by copying the default channel group so all the standard channels remain.

Step 2: Add an AI assistants channel

Add a new channel named "AI assistants" with a condition on Session source that matches regex:

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

Adjust the list for the client's markets. Some regions have popular local assistants worth adding.

Step 3: Order it above Referral

Channels are evaluated in order, and a session is assigned to the first one that matches. If Referral comes first, AI visits will be captured there. Move "AI assistants" above Referral.

Step 4: Use it in reports

Select the new channel group in acquisition reports or explorations. Because custom channel groups apply to reporting rather than to collection, they can classify historical sessions too, which makes before-and-after comparisons possible.

Step 5: Repeat the setup per property

Agencies managing many properties should document the regex in one place and apply the same definition everywhere, so AI traffic is comparable across clients.

Measuring the Direct traffic gap

Some AI visits will always arrive as Direct. Estimates of the share vary widely, and some sources suggest that a large portion of AI-assisted visits are recorded without a source. You cannot recover those visits exactly, but you can watch for signs of them.

• Landing pages. A rise in Direct sessions landing on deep pages, such as specific guides or product pages, is unusual for genuine direct visits. People rarely type long URLs. If those pages are also cited in AI answers, AI is the likely source.
• Timing. Look for Direct spikes on pages shortly after they start being cited.
• Engagement. Visits from AI answers often show distinctive engagement, for example longer sessions on a single page.

Report this as a likely range, not a precise figure. Clients trust estimates they can see the reasoning for.

Building an exploration for AI traffic

The standard acquisition reports are enough for a headline number. An exploration gives you the detail clients ask about next.

1. In GA4, open Explore and start a Free form exploration.
2. Add the dimensions Session source, Landing page + query string, and your custom channel group.
3. Add the metrics Sessions, Engaged sessions, Key events (conversions) and, for e-commerce, Total revenue.
4. Set a filter so the channel group equals "AI assistants".
5. Put the channel and source in rows and the metrics in values; add landing page as a second row dimension.

This shows which assistants send visits, which pages they land on and whether those visits convert. Save it as a template and copy it to each client property so every report uses the same view.

For a quick check of the Direct gap, build a second exploration filtered to the Direct channel, with landing page as the row dimension. Deep pages with unusually high Direct sessions are the ones most likely to be receiving AI visits without a referrer.

What to report to clients

A compact AI traffic section in the monthly report works well.

Pair it with visibility data: how often the client is mentioned and cited in AI answers for their prompt set. Traffic shows the clicks; visibility shows the exposure that did not produce a click. See AI share of voice.

Interpreting the numbers

AI referral traffic is usually small compared with organic search. That is normal in 2026, and it should not be the headline. Three points help clients read it correctly.

1. The trend matters more than the total. Month-on-month growth in AI sessions, especially to commercial pages, is the useful signal.
2. Quality often outweighs volume. Visitors who arrive after reading an AI answer have often already narrowed their choice. Compare conversion rates from AI sessions with organic sessions.
3. Platform shifts can move the numbers suddenly. Changes to how an assistant formats links can change referral volumes overnight. One monitoring service reported a sharp drop in ChatGPT referral traffic across the sites it tracked in October 2026 while total traffic held steady. When AI traffic moves sharply across many clients at once, suspect a platform change before blaming the site.

Concentration across assistants

Studies of AI referral traffic consistently find that ChatGPT sends the large majority of trackable AI visits, though estimates of its share range from roughly three-quarters to more than 90%, with Perplexity usually second. That concentration varies by audience. B2B and technical audiences often show more Perplexity and Claude traffic than consumer audiences. Report by assistant, not just in total, so clients see where their audience actually is.

Tagging links you control

You cannot add UTM parameters to links AI engines generate, but you can tag links you control that appear in AI contexts. Examples include links in a custom GPT, an app listing in an AI marketplace, or links in documentation that AI tools commonly fetch. Use a consistent convention such as utm_source=chatgpt&utm_medium=ai-assistant, and make sure your channel rules catch it.

Common mistakes

• Leaving AI visits in Referral. They get lost among ordinary referral sites.
• Putting the AI channel below Referral. It will never match.
• Counting Google AI Overviews as AI assistant traffic. Those clicks are organic search.
• Claiming precise totals. Some AI traffic is unattributable. Present ranges where needed.
• Different definitions per client. Keep one regex and one channel name across the portfolio.

How Rankwyn helps

Rankwyn pulls each client's AI assistant sessions into the same report as their AI citations from the AEO Citation Radar, so clients see exposure and visits side by side, and white-label multi-site reports apply the same definitions across every property you manage.

Key takeaways

• Without setup, GA4 splits AI visits between Referral and Direct and labels none of them as AI.
• Create a custom channel group with an "AI assistants" channel using a source regex, and place it above Referral.
• Use GA4's built-in AI Assistant channel where it exists, but keep the custom group for history and consistency.
• Some AI visits arrive as Direct; estimate them using deep-page Direct landings rather than claiming precision.
• Google AI Overviews and AI Mode clicks are organic search, not AI assistant traffic.

The bottom line

AI assistant traffic is measurable enough to report, as long as you set GA4 up deliberately and are honest about the gaps. Build one AI channel definition, apply it across every client property, report sessions, landing pages and conversions by assistant, and pair the numbers with citation data so clients see the whole picture. For the wider measurement approach, see proving SEO value when clicks fall.