How to Measure AI Referral Traffic in GA4 (Setup and Limits)
Short answer: in GA4, check whether the native “AI Assistant” channel is in your default channel group, then create a custom channel group with a source regex for AI engine domains, placed above Referral, so you can see these visits as one channel. Treat the result as a floor. Referrers get stripped, some AI traffic lands in Direct, and clicks from Google’s AI features are counted as Organic Search.
I verified the setup steps against reports from industry publishers in November 2026, and I flag where they disagree. GA4’s interface and Google’s channel definitions change, so confirm each step in your own property.
- AI referrals show up in GA4 as sessions with sources like chatgpt.com or perplexity.ai, when the engine sends a referrer.
- Google added a native AI Assistant channel in 2026, but it does not capture everything.
- A custom channel group with a regex gives you control and consistent reporting.
- Referral numbers undercount AI influence. Use them alongside prompt tracking.
What AI referral traffic is, and what it is not
When someone clicks a link in a ChatGPT, Perplexity, Gemini or Copilot answer, the browser can send the engine’s domain as the referrer. GA4 records that as the session source. That is AI referral traffic.
It is not:
- Everyone who saw your brand in an answer. Most readers of an AI answer never click. They remember a name and come back later through search or directly.
- Clicks from Google’s AI Overviews or AI Mode. Those arrive as Google search traffic. Google says traffic from its AI features is included in the Web search type in Search Console and, in GA4, it counts as Organic Search.
- Traffic you can always attribute. Some AI apps and in-app browsers strip the referrer, and those visits land in Direct.
So the number you get is real but partial. Use it to learn which engines send people, which pages they land on and whether those visitors convert, not to measure how visible you are.
Step 1: See what you already have
Before building anything, look at the data you have.
- In GA4, open Reports > Acquisition > Traffic acquisition.
- Change the primary dimension to Session source / medium.
- Search the table for chatgpt.com, perplexity, gemini.google.com, copilot.microsoft.com, claude.ai and similar.
You will probably find a few of them, split across mediums such as referral and (not set). That split is the first sign of why a custom view helps.
ChatGPT links often carry a utm_source=chatgpt.com parameter, according to several practitioner guides. If that is present, GA4 may record the source from the UTM rather than the referrer. I have not found Google or OpenAI documenting this, so check what your own data shows rather than relying on it.
Step 2: Check the native AI Assistant channel
According to industry reports, Google added an “AI Assistant” channel to the default channel group on May 13, 2026, rolling out gradually and reaching most properties by early June. When GA4 recognizes a listed AI referrer, it tags the session with the medium ai-assistant and the campaign (ai-assistant).
Things to know, and the places reports disagree:
- Coverage. At launch the reported list included ChatGPT, Gemini and Claude. A later report listed ChatGPT, Gemini, DeepSeek, Copilot and Grok, and said Perplexity is not on it. Sources are inconsistent, so treat the list as unverified and compare it with your own sources.
- Only from rollout. The channel counts tagged sessions from when it reached your property. Comparing periods that span the rollout mixes tagged and untagged data.
- Split across channels. The same source, such as chatgpt.com, can appear in AI Assistant, Referral or Unassigned depending on the medium. This is the main reason it undercounts.
- You cannot edit the default channel group. To change the rules you need a custom group.
To check it in your property: open Reports > Acquisition > Traffic acquisition, set the dimension to Session default channel group, and look for “AI Assistant.” If it is there, click into it and see which sources it contains.
Step 3: Build a custom channel group
This is the most reliable way to get one consistent AI channel. The steps, which match several published guides:
- Go to Admin > Data display > Channel groups.
- Choose Create new channel group and name it, for example “Default plus AI.”
- Add a new channel, named “AI.”
- Set the condition to Source matches regex and enter your pattern (below).
- Drag the AI channel above Referral and Organic Search so it claims those sessions first. GA4 evaluates channels top to bottom.
- Save, then use the group as a primary dimension in your acquisition reports.
If you copy the default channel definitions into the new group, do it before adding AI so your other channels keep working. Check the interface wording in your own property, since Google moves menu items.
The regex
Keep it anchored to domains so you do not catch unrelated traffic. A simple pattern that covers the common AI engines:
(^|\.)(chatgpt\.com|chat\.openai\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|copilot\.microsoft\.com|grok\.com|deepseek\.com)$
Notes on building yours:
- Anchor the ends. Without anchors, a loose pattern can match unrelated domains.
- Avoid bare words. A pattern containing just “gpt” or “google” produces false positives. One report specifically warns that Google’s own sample regex is too loose to use as written.
- Be careful with shared domains. Adding bing.com would pull in ordinary Bing search referrals, so leave it out unless you want that.
- Check your own sources first. Build the pattern from the AI sources you actually see in Step 1, then add others you want to catch in future. Review the list every quarter, since domains and products change.
- Test it. In Explore or the traffic acquisition report, confirm that sessions you know are from ChatGPT or Perplexity are landing in the AI channel.
One guide says GA4 allows a limited number of custom channel groups per property, so check what you already have before creating another.
Step 4: Decide how to treat history
Sources disagree on whether a custom channel group reclassifies past data. Some say it does, since custom channel groups are applied at report time, and one advises validating it. Settle it in your property: after saving the group, open a date range from before today and see whether the AI channel has sessions.
Whatever the answer, note in your reporting the date you created the group and the date the native channel arrived. Without that note, a jump in “AI traffic” in June 2026 is hard to interpret.
Step 5: Build the reports that matter
With the channel in place, three reports earn their keep:
Which engines send traffic. Session source within the AI channel. Expect ChatGPT and Perplexity to dominate in many accounts, though this varies by audience.
Which pages they land on. Use landing page as the dimension. This tells you which of your pages engines are linking, which is the nearest thing GA4 gives you to a citation signal. Compare it with the prompts you track.
Whether it converts. Use key events or conversions as metrics. Many teams find AI referrals are small in volume. Whether they convert better than average depends on your business, and you should test that on your own data rather than assume it.
Add a comparison with the same period last year or last quarter, and keep the caveat about the pre-June tagging in mind.
Step 6: Fill the gaps with links you control
For links you place yourself, such as your bio on a community, a guest post or a partner page, use UTM parameters so the visits are attributed regardless of what the app sends. Do not tag links that a person might share in a way that looks like manipulation, and do not put UTMs on links meant to rank in search, because the parameters can create duplicate URLs.
For campaigns that ask people to find you through an AI engine, a simple “how did you hear about us” field on signup catches the remainder. A free-text field with an “AI assistant” option often reveals more than analytics does.
Limits you should state in every report
- Stripped referrers. Some AI traffic arrives with no referrer and appears in Direct.
- Zero-click influence. Readers who see your name and search for it later look like branded organic or direct traffic.
- Google’s AI features. Counted as Organic Search, with no separate GA4 breakdown.
- Definition changes. The native channel’s list of platforms has changed since launch, per reports.
- Small numbers. For many sites AI referral volume is a small share of traffic, so a few visits can swing percentages. Report counts as well as rates.
- Bots and previews. Some fetches by user-triggered bots are not human visits. GA4 filters known bots, though not perfectly. Server logs can help separate them.
Pair it with prompt tracking
Referral traffic answers “who clicked?” It does not answer “are we showing up?” For that you need prompt-level tracking: a fixed list of buyer questions, run repeatedly, with mentions and citations recorded. The two together tell the real story. If landing-page referrals rise on pages that you recently got cited for, you have a causal story. If mentions rise and referrals do not, the answers may be satisfying readers without a click, which is still worth something.
How to track brand mentions in AI search covers the prompt side, and AI search impressions are a vanity metric explains why to be careful with modeled numbers. ReddGrow’s AI brand visibility tracking handles prompts, citations and share of voice if you do not want to maintain a spreadsheet.
Checklist
- Look at Session source / medium for AI domains.
- Confirm whether the AI Assistant channel exists and what it contains.
- Create a custom channel group with an anchored regex, above Referral.
- Test with known sessions and a past date range.
- Report AI traffic with the caveats above and the dates of each change.
- Pair it with prompt-level tracking.
Frequently Asked Questions
- Does GA4 track traffic from ChatGPT automatically?
- Partly. Industry reports say Google added a native AI Assistant channel to GA4's default channel group in May 2026, covering a list of recognized AI referrers. Reports disagree on which platforms are on that list, and it does not catch everything, so verify it in your own property and consider a custom channel group.
- Why does some AI traffic show up as Direct?
- Some AI apps and in-app browsers do not pass a referrer, so GA4 has nothing to attribute the visit to and files it under Direct. That means any AI referral number is a floor, not a total. You cannot fully fix it, though tagged links you control can help.
- Are Google AI Overviews and AI Mode counted as AI traffic in GA4?
- No. Clicks from Google's AI features arrive as Google search traffic and are counted as Organic Search in GA4. Google also says traffic from AI features is included in the Web search type in Search Console, without a separate breakdown you can filter on.
- Will a custom channel group change my historical data?
- Sources disagree, with some saying custom channel groups apply retroactively and others advising you to confirm. Test it in your own property by checking a past date range after creating the group. Do not assume it either way when you report trends.
- Is AI referral traffic a good measure of AI visibility?
- It is a useful business outcome but a weak visibility measure. Many people read an AI answer, remember a brand and search for it later, which shows up as direct or organic traffic. Pair referral data with prompt-level tracking of mentions and citations.