GA4 AI Referral Traffic: A Hands-On Setup Guide

GA4 AI referral traffic can be separated from ordinary referrals with a custom channel group and rules that match known AI assistant sources. This setup gives you a clearer view of visits from tools such as ChatGPT, Perplexity, Claude, Gemini and Copilot, including historic traffic that the default group may not classify as AI. It cannot recover visits where the browser or app did not send a referrer.

The goal is a dedicated channel you can use in GA4 reports and Explorations, without changing the existing Default Channel Group. If you also want to understand the difference between AI citations and the visits they send, see our guide to AI search visibility and content strategy.

What this setup does, and what it cannot do

A channel group is a collection of rules that classifies sessions into channels such as Organic Search, Referral and Direct. GA4’s Default Channel Group uses Google’s rules. You can’t edit or reorder those rules, but you can create a Custom Channel Group with your own rules and order. Google explains the difference in its guide to default channel groups and its Custom Channel Groups documentation.

The custom group in this tutorial adds an AI Referral channel. Its rules look for AI-related source names, referrer URLs and GA4’s native AI Assistant channel where that information is available. Similar channel configurations are detailed in Growth-onomics' guide to the GA4 AI assistant channel. We’ll place it above Referral so an AI source does not get caught by the more general rule first.

There are important limits:

• GA4 only records visits where its tag runs. A crawler that reads your site without loading the page’s GA4 tag will not appear as a visitor.
• Some AI apps and browser handoffs remove referral information. Those visits may appear as Direct or Unassigned, and a channel rule cannot identify their true source if GA4 did not receive it.
• Google’s native AI Assistant channel is not a complete historical record. Google announced the channel on 13 May 2026. Its introduction does not recategorise earlier visits, and it excludes Google AI Overviews and AI Mode, whose clicks remain under Google organic traffic. See Google’s Analytics update.
• AI visibility is not the same as referral traffic. GA4 can report visits and on-site activity, not how often an AI assistant mentioned your brand without a click.

Treat the new channel as a useful reporting view, not a complete measure of AI discovery. For a broader view of visibility and performance, SearchFIT’s resources cover AI search measurement alongside site traffic.

Before you start

What you need

Prepare these items before editing the property:

• Access to the correct GA4 property, with permission to create or edit channel groups. If the channel group controls are unavailable, ask a property administrator to check your access.
• A list of AI assistant sources relevant to your audience. The rule below includes ChatGPT, Perplexity, Claude, Gemini, Copilot and several other commonly recognised AI products. You can add or remove names to suit your reporting needs.
• A note of the reports you use to review acquisition, engagement and conversions. After setup, you will need to select the custom channel group as a reporting dimension.
• A test plan: one or more real visits from AI tools if possible, plus a date range that includes historic sessions.

You do not need to install a new tracking tag for this channel group. It uses data GA4 already receives. It also does not change your campaigns, source and medium values, or your Default Channel Group.

Choose a useful name

Use a name that will still make sense to other people in your account. This tutorial uses Primary Channels (with AI) for the group and AI Referral for the new channel. If your team already has a channel naming convention, follow it. Avoid naming the whole group AI Referral: the group includes your other channels as well.

Set up GA4 AI referral traffic as a custom channel

Step 1: Open Channel groups

1. Sign in to Google Analytics and select the correct property.
2. Open Admin using the gear icon.
3. Under Data display, open Channel groups.
4. Select Create new channel group.
5. Choose to copy or duplicate the Default Channel Group if that option is available in your interface. This gives you a familiar starting point while leaving the original group unchanged.
6. Name the group Primary Channels (with AI).

What you should see: a channel group editor containing the channels copied from the default group, with controls to add a channel and change the order. If you start from a blank group instead, you will need to build all the channel rules you want in the group. Duplicating the default is usually simpler for this tutorial.

Google’s interface can change. If the menu labels differ, look for the property’s channel group settings under Admin and Data display. The Google help page for custom channel groups describes the feature and its purpose.

Step 2: Add a dedicated AI channel

1. In the group editor, select Add new channel.
2. Name the channel AI Referral.
3. Open its rule editor and add conditions that identify AI sources.
4. Build the alternatives as separate condition groups joined with OR. Do not put every alternative into one group joined with AND. An AND rule would ask one session to satisfy every condition at once, which is not the intention.

Use these three alternatives:

Condition group A: Source matches a regular expression

Choose the Source dimension and the matches regex operator. Add this expression:

.*(chatgpt|openai|perplexity|claude|anthropic|gemini|copilot|meta\.ai|searchgpt|deepseek|grok|phind|you\.com|poe\.com|mistral).*

This checks the source value for one of the listed names. The expression uses .* before and after the list so it can match a name within a longer value. The backslash before the dot in names such as meta.ai and you.com means “a literal full stop”, rather than any character.

Condition group B: Page referrer matches a regular expression

Add another condition group using the Page referrer dimension and matches regex:

.*(chatgpt\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|copilot\.microsoft\.com|android-app:\/\/com\.openai\.chatgpt).*

This looks for known assistant domains in the referrer URL, including an Android app referrer pattern. The pattern is a catch for values GA4 may receive; it does not mean every mobile app click will send one. Referrer data can be absent or stripped during the handoff to a browser.

Condition group C: Default channel group exactly matches the native AI channel

Add a third condition group using Default channel group and the exactly matches operator. Enter AI Assistant as the value.

This alternative catches sessions Google has already put into its native AI Assistant channel. The condition name and available dimensions can vary slightly in the interface. If you cannot select Default channel group in this rule editor, leave this alternative out and use the source and referrer rules. Do not replace it with an unrelated dimension.

What you should see: three separate ways for a session to qualify for the AI Referral channel. A session only needs to meet one of the alternatives. Check the editor’s connectors before saving: the condition groups must be combined with OR.

Step 3: Put AI Referral above Referral

Rule order matters. GA4 evaluates channel rules from top to bottom, and a session is assigned according to the first matching rule. If the generic Referral rule is above AI Referral, a session with perplexity.ai / referral can be classified as Referral before the AI-specific rule gets a chance to match.

1. Select Reorder in the channel group editor.
2. Move AI Referral above Referral.
3. Keep it above any other broad channel rule that could match the same sessions, including Organic Search if your rules could classify an AI source that way.
4. Review the remaining channel order. Keep the standard channels copied from the default group unless you have a deliberate reason to change them.
5. Select Apply or the equivalent order-confirmation control.

What you should see: AI Referral positioned before Referral in the group’s channel list. The order of channels in this custom group does not change the Default Channel Group.

Step 4: Save the custom group

1. Review the group name and the AI Referral rule conditions.
2. Confirm the AI channel sits above Referral.
3. Select Save channel group.
4. Return to Admin > Data display > Channel groups and confirm your new group appears in the list.

If GA4 displays a warning about rule order or a condition that is not recognised, fix that before saving. A saved channel group with a misplaced rule can be less useful than no new channel at all, because it may look like AI traffic is still being categorised as ordinary referral traffic.

Check whether the setup worked

Do not rely on the channel group appearing in Admin as proof that it is classifying sessions correctly. Check the group in reports, test the rule logic and compare the results with the source and medium values GA4 received.

Check historic traffic in a report

1. Open a standard acquisition report, such as Traffic acquisition.
2. Change the primary dimension to the channel group you created. Depending on the report, it may appear as the name of your custom group or as a custom channel group dimension.
3. Set a date range that includes the period you want to inspect.
4. Look for AI Referral in the channel rows.
5. Add or switch to source and medium as a secondary dimension if the report allows it. Review which source names are contributing to the channel.

What you should see: an AI Referral row if the property has sessions matching your conditions during the selected period. Check the source breakdown for expected names such as chatgpt.com, perplexity.ai or another listed source. A source may use a different value, so update your expression if you see a relevant AI source outside the channel.

Custom channel groups can be applied to historical data in reports. This differs from changing a source value after collection: you are asking GA4 to classify the existing data with your new group’s rules. Allow time for the reports to reflect the change. As noted in Cognizo's guide to AI traffic in GA4, standard reports and Explorations may take 24 to 48 hours to recalculate; the actual availability can vary by report and property.

Use an Exploration to inspect the sessions

An Exploration gives you more control over the dimensions and metrics you compare.

1. Open Explore and start a free-form exploration.
2. Add the custom channel group dimension, along with Session source / medium and Landing page + query string if available in your property.
3. Add metrics that answer your question, such as sessions, engaged sessions, key events and total revenue where relevant.
4. Put the channel group in rows and the metrics in values. Add source and medium as a breakdown or use a filter for AI Referral.
5. Select the same date range you used in your acquisition report.

What you should see: sessions grouped under AI Referral, with source and landing page detail that helps you understand who sent them and where they arrived. If the dimension is not available in the Exploration yet, check again after the group has had time to process. Also check that the Exploration’s date range includes matching traffic and that the report is using your custom group, not the Default Channel Group.

Compare with source and medium

For a quick sense check, compare three views over the same date range:

The totals will not necessarily match one for one. Channel grouping rules, attribution settings, date ranges and the available source information affect what you see. Make sure you are comparing session-based dimensions with session metrics, rather than mixing session and event scopes without checking what each metric means.

Test a new visit

A live test can confirm that your site is collecting data, but it may not immediately show the custom channel in every report.

1. Open an AI assistant in a browser and use a real link to your site, if one is available. Avoid editing the link in a way that changes its referrer behaviour.
2. Open your site and check that the GA4 tag fires using your normal consent and debugging tools.
3. In GA4, inspect Realtime or DebugView to confirm that activity reached the property. Check the available source and referrer details where shown.
4. After the reporting delay, review the session in the acquisition report using your custom channel group.

What you should see: the visit reaches GA4, and the report classifies it as AI Referral if its source or referrer matches your conditions. If Realtime shows the visit but the custom report does not, first check the date range and report dimension, then inspect the source and referrer values. The channel group cannot match a value that was never sent.

Make the regex fit your property

The expressions above are a practical starting point, not a universal directory of every assistant, app and referral format. Review the actual values in your GA4 property before adding more names. You can inspect source values in Traffic acquisition or an Exploration, then add a name when it represents an AI tool you want in this channel.

Keep these rules in mind:

• Use the values GA4 records. A product brand and a referral hostname are not always the same string. Use source and referrer values from your own reports to guide changes.
• Avoid matching vague words alone. A broad term such as ai can catch unrelated sources. Specific product names and domains reduce accidental matches.
• Escape dots in domains. In a regular expression, a dot normally stands for any character. Use \. when you mean a literal dot in a hostname.
• Review new sources periodically. AI products and their referral behaviour can change. Check the source breakdown after releases or when you notice unfamiliar traffic.
• Keep a record of edits. Write down the date, expression and reason for each change. This makes it easier to explain why historic totals may have shifted.

If a source uses a new value, add it carefully rather than replacing the whole expression. Then check that the new alternative still uses OR logic and that AI Referral remains above Referral.

Interpret the numbers without overclaiming

Once you can isolate AI referrals, assess what those visitors do on your site. Useful questions include:

• Which AI sources bring sessions to the site?
• Which landing pages receive those visits?
• Do those sessions engage with the page or trigger a key event?
• Which visits contribute to leads, purchases or other business outcomes?
• Are the results changing across comparable date ranges?

Use GA4’s own event and conversion configuration consistently. A channel group labels traffic; it does not define what counts as a lead, purchase or qualified opportunity. Check that the key events you compare are configured correctly and that revenue values are being collected as intended.

Also separate AI referrals from AI visibility. A visitor who clicks a citation can appear in GA4 if tracking runs and a source is passed. A brand mention that does not lead to a click cannot appear as a GA4 session. Google routes clicks from AI Overviews and AI Mode through google / organic, rather than the native AI Assistant channel, so GA4 does not provide a simple standard channel that isolates those experiences. Our guide to AEO KPIs explains why referral sessions are only one part of evaluating AI search performance.

Finally, do not interpret a drop in a named AI source as proof that the assistant stopped mentioning your brand. It may reflect fewer clicks, a change in referrer handling, or incomplete source data. Compare GA4 with your other measurement and visibility data, and make the limits clear when reporting results.

Common problems and how to fix them

AI sessions still appear under Referral

Check that AI Referral sits above Referral. Confirm that the report uses your custom channel group rather than the Default Channel Group. Then inspect the actual source value: if it is not included in your regex, add the specific source name and test again.

Perplexity does not appear in the native AI channel

The native AI Assistant channel does not include every AI source. Perplexity can appear as a standard referral instead. The custom source and referrer rules in this guide are intended to catch it when GA4 receives a matching value.

ChatGPT traffic appears as Direct or Unassigned

A referrer may be missing when a user moves from an app or a restricted web view to a browser. GA4 cannot infer the referring assistant from a session with no identifying information. A custom channel group will not recover that lost information. You can still monitor Direct and Unassigned for unusual changes, but avoid relabelling all such traffic as AI.

There are no historical sessions in AI Referral

Check whether the property actually recorded matching source or referrer values in the selected period. Confirm the group was saved, the conditions use OR, and the report uses the custom group. Then allow time for reporting to update. If none of the historic source values match, the new channel will not create sessions that were not present in the data.

Organic Search looks unchanged

That may be expected. Google AI Overviews and AI Mode clicks remain under Google organic traffic in standard GA4 reporting, and their visits are not separated by the AI Assistant channel. A custom regex based on assistant hostnames does not by itself split those Google clicks from other organic visits.

How to keep the channel useful

Review the setup when your reporting needs change, not every time you see a small movement in traffic. A monthly or quarterly check is a reasonable operating habit, but choose a cadence that fits your reporting cycle.

At each review:

1. Check for new AI-related source values in acquisition reports.
2. Confirm the main known sources still match the expression.
3. Test that the AI Referral rule remains above Referral.
4. Compare channel sessions with source and medium for the same dates.
5. Note any changes to the regex, report setup or tracking implementation.
6. Explain missing referrers and Google AI search limitations in reports to stakeholders.

If you need to evaluate AI search beyond sessions that click through, use a separate visibility measurement approach. SearchFIT’s pricing and platform information can help you assess whether a dedicated AI visibility and search measurement workflow fits your team. GA4 and AI visibility tracking answer related, but different, questions.

Next step

Create the custom group, confirm its rule order and test it against real source values from your property. Then use AI Referral alongside source, landing page and key event data to understand which visits matter, while reporting untracked mentions and missing referrers as limitations rather than filling in the gaps with guesses. If you want to evaluate how an automated visibility platform compares to traditional manual monitoring, see SearchFIT vs agencies for a side-by-side breakdown.