How to See AI Referral Traffic in GA4 (Step by Step)

To see AI referral traffic in GA4, filter your sessions by the source domains of the AI assistants: chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, claude.ai and similar. You can do it in three ways, from a 2-minute report filter to a permanent custom channel group, and each has a limit you should know before trusting the number.

This guide covers GA4 only. It measures visits, which is the outcome layer of AI search measurement; it does not show answers that mention you without a click.

What AI traffic does GA4 already classify?

Google’s default channel group now contains an AI Assistant channel. Per Google’s default channel group documentation, it covers visits from sources like ChatGPT, Gemini, Deepseek, Copilot and Grok, and it excludes Google’s AI Overviews and AI Mode, which are counted under Organic Search. The documentation does not list every source in the match list. That is why many teams still add their own regex: you can see exactly what is included and add assistants as they appear.

Check first: Reports > Acquisition > Traffic acquisition, and look for “AI Assistant” in the Session default channel group. If it shows traffic, you already have a baseline.

Method 1: Filter Traffic acquisition by source (2 minutes)

1. Go to Reports > Acquisition > Traffic acquisition.
2. Change the first column dimension to Session source / medium using the dropdown above the table.
3. Click the search box and enter chatgpt.com, then repeat for perplexity.ai, gemini.google.com, copilot.microsoft.com and claude.ai.
4. Set the date range to at least 28 days.

This is quick but manual, and you cannot see the total in one row.

Method 2: Build a free-form exploration with a regex (10 minutes)

Explorations let you filter on several sources at once.

1. Open Explore and create a Blank or Free form exploration.
2. In Variables, import the dimension Session source and the metrics Sessions, Engaged sessions, Key events and Total revenue (if you track revenue).
3. Double-click Session source to add it to Rows, and drag the metrics to Values.
4. Under Filters, add a filter on Session source, set the match type to matches regex, and paste:

chatgpt\.com|openai\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|claude\.ai|chat\.deepseek\.com|grok\.com

5. Optionally add Landing page + query string as a second row dimension to see which pages the assistants send people to.

Google’s documentation says an exploration supports up to 10 filters per tab. Save the exploration and name it “AI referrals”.

Method 3: Create a custom channel group (permanent, 15 minutes)

A custom channel group puts AI traffic into its own channel in every standard report.

1. Go to Admin > Data display > Channel groups. You need Editor access or above at the property level.
2. Copy the default channel group (the predefined one cannot be edited) and name the copy, for example “Default plus AI”.
3. Add a new channel called AI referrals with the condition Source matches regex and the same pattern as above.
4. Move the new channel above Referral and above any channel that would otherwise capture it. Traffic goes into the first channel whose conditions it matches.
5. Apply the group. You can then make it your primary channel group from the same page.

Limits per Google’s custom channel groups documentation: standard properties can have 2 custom groups and each group up to 50 channels. Reports can apply a new group retroactively, but audiences only apply it going forward.

You can also create a segment in an exploration with the same regex condition on Session source to compare AI visitors with everyone else on engagement and conversion.

What should you look at once AI traffic is visible?

Question	Where to look	What it tells you
How much is there?	Sessions by AI source	Scale and trend
Is it good traffic?	Engagement rate, key events per session vs. organic	Quality of AI visitors
Which pages receive it?	Landing page	Pages engines cite and link
Which assistant sends it?	Session source	Where to focus

Compare the landing pages with the pages you want cited. If AI visits land on old blog posts rather than your product or pricing pages, that is a content gap you can act on.

Why does GA4 undercount AI traffic?

Four known causes:

1. Missing referrers. Some assistants and mobile or in-app browsers open links without passing the referring domain, so GA4 records the visit as Direct.
2. Copy and paste. A user who copies your URL from an answer into a new tab arrives as Direct.
3. Mixed channels. Clicks from Google AI Overviews and AI Mode land in Organic Search with no separate label.
4. Zero-click exposure. People who read the answer, then search for your name later, appear as branded organic or direct visits.

So GA4 gives a floor. Directionally, a rising AI line is a good sign; a low one does not mean low exposure. Watch Direct and branded search alongside it, as covered in zero-click AI search.

How do you tag links you control?

If you share links where assistants may pick them up (documentation, partner pages, press kits), nothing stops you tagging with UTM parameters, though you cannot tag the links an engine chooses to cite. Some assistants append a source parameter themselves; check your Landing page + query string report for it rather than assuming.

How does server-side tracking complement GA4?

GA4 is client-side JavaScript and only sees visitors who load it. AI engines also fetch your pages directly, with no browser and no GA4 hit. A tracking script or server-side log analysis can see both. GeoGen’s LLM traffic feature records AI crawler visits and AI referral visits through a tracking script and is included on every plan. It answers “which engines fetch which pages”, which GA4 cannot, while GA4 remains the better place for conversion analysis. Use both: server-side for what engines read, GA4 for what visitors do.

To pair this with the answer-level view (are you named, cited, described well), see how to rank in ChatGPT and the stakeholder framing in AI search ROI. Plans start at €25 per month on the pricing page.