Google Analytics 4 for GEO: 11 Ways to Measure (and Grow) Your AI Traffic

Lee Stacey, Bourne Digital · September 2026

Search Console tells you when AI shows your content.

Google Analytics tells you what happens after someone clicks.

For a B2B company, that second part is the one the board cares about. Did the visitor from ChatGPT read three pages and book a demo? Or bounce in four seconds?

The good news: in 2026, GA4 finally started treating AI traffic as its own thing.

The bad news: it still misses a lot.

In this guide I'll show you:

• What GA4 can (and can't) see from AI search
• How to set it up so you catch as much AI traffic as possible
• How to use that data to decide where to focus your GEO efforts

What GA4 Can and Can't See

Before we touch any settings, you need to know where AI traffic actually ends up in GA4.

It's split across four different places:

And one more that GA4 can't see at all:

AI answers that mention you without anyone clicking. No click, no session, no data.

So keep this in mind for the rest of this guide:

GA4 AI traffic numbers are a floor, not a ceiling. The real figure is higher.

With that said, here's how to get the most out of it.

Part 1: Set Up

#1: Find the new AI Assistant channel

On 13 May 2026, Google Analytics added a new default channel called AI Assistant. (Google Analytics Help)

Google describes it as “the channel by which users arrive at your site from sources like ChatGPT, Gemini, Deepseek, Copilot, or Grok.” (Google Analytics Help)

Here's how it works. If the referrer matches Google's list of AI assistants, GA4 sets:

• Medium to ai-assistant
• Campaign to (ai-assistant)

and the session lands in the AI Assistant channel instead of Referral. (Google Analytics Help)

To see it, go to:

Reports → Acquisition → Traffic acquisition

Make sure the table is showing the default channel group dimension and look for AI Assistant.

Then change the dimension to Session source / medium and filter for ai-assistant to see which AI tools are sending you traffic.

Two important catches:

1. It isn't retroactive. Sessions from before 13 May 2026 stay in whatever channel they were in (usually Referral).
2. Google doesn't publish the full list of AI sources. The examples in the help docs have changed since launch. So don't assume a tool is (or isn't) covered. Check your own data.

#2: Build a custom AI channel group to catch the rest

The default channel doesn't catch every AI tool, and it doesn't cover older data.

A custom channel group fixes both.

Why? Because Google says “custom channel groups can be applied to your reports retroactively.” (Google Analytics Help) So you can see your AI traffic trend going back well before May 2026.

Here's how to set it up:

Step 1: Go to Admin → Data display → Channel groups. You'll need Editor access or above.

Step 2: Click Create new channel group. GA4 copies the default channels in for you.

Step 3: Click Add new channel and name it AI Search. Set the condition to:

Source → matches regex →

.*(chatgpt|openai|perplexity|claude|gemini\.google|copilot|deepseek|grok|meta\.ai|mistral).*

Step 4: This is the step most people get wrong. Drag the AI Search channel above Referral.

Google says traffic goes into “the first channel whose definition it matches given the current order of channels in the group.” (Google Analytics Help) If Referral sits above your AI channel, Referral catches the traffic first.

Step 5: Save. Then in any acquisition report, switch the channel dimension to your new group.

Pro Tip: Standard GA4 properties only get two custom channel groups (GA4 360 gets five). (Google Analytics Help) Don't waste one. Build one master group that includes AI Search and any other channel tweaks you need.

Pro Tip #2: Every few months, check Referral for new AI tools and add them to the regex. New ones launch all the time.

#3: Check for ChatGPT traffic hiding in “Unassigned”

ChatGPT often adds utm_source=chatgpt.com to the links it cites.

That's helpful, but a UTM source with no medium can sometimes end up somewhere unexpected.

So do a quick check:

Reports → Acquisition → Traffic acquisition → Session source / medium, then search for chatgpt.

If you see chatgpt.com / (not set) or ChatGPT sessions in Unassigned, your custom channel from #2 will pick those up too, because it matches on source, not medium.

#4: Mark your B2B conversions as key events

This one isn't AI-specific. But without it, everything else in this guide is just traffic counting.

In B2B, AI visitors rarely buy on the spot. So make sure GA4 is tracking the actions that actually lead to pipeline:

• Demo or consultation requests
• Contact form submissions
• Pricing page views
• Case study or whitepaper downloads
• Newsletter sign-ups

Mark the important ones as key events (Admin → Data display → Events).

Now you can answer the question that matters: does AI traffic turn into leads?

Part 2: Analyse

#5: Find the pages AI sends people to

Go to Reports → Engagement → Landing page.

Add a filter (or comparison) for your AI Search channel.

This shows you exactly which pages AI tools are linking to.

Now put this list next to the Pages list from Search Console's Generative AI report (see our Search Console GEO guide).

You'll get three groups:

That third group is interesting. The page is being shown, but the snippet or title isn't making people want to click. That's often a sign the AI answer covered everything, and your page didn't promise anything extra.

#6: Compare AI traffic quality with organic

Now for the question everyone wants answered.

In Traffic acquisition, using your custom channel group, compare AI Search with Organic Search on:

• Engagement rate
• Average engagement time per session
• Key events and session key event rate

Don't assume AI traffic is automatically better or worse. Check.

AI visitors may arrive with more context, because they've already had the “what is X” conversation with the AI. Or they may just be checking a single fact. It varies a lot by business. Your data is the only data that matters here.

Pro Tip: AI traffic is usually small. Look at 3–6 months of data before drawing conclusions. A handful of sessions can make the percentages swing wildly.

#7: Follow the AI visitor's journey with Path exploration

Go to Explore → Path exploration.

Create a segment of sessions from your AI Search channel and apply it. Then set the starting point to Page path and pick one of your top AI landing pages from #5.

This shows you where AI visitors go next after landing.

Look for:

• Do they reach pricing or contact pages? Good. Your content is doing its job.
• Do they leave after one page? Maybe the page answered their question fully (which isn't always bad). Or maybe there's no clear next step.
• Do they bounce back to a page the AI already summarised? That may mean your page didn't add anything new.

Then fix the journey. For example: add a clear “see how this works for [industry]” link, a relevant case study, or a short product demo video on your top AI landing pages.

#8: Link Search Console to GA4

This one takes two minutes.

Go to Admin → Product links → Search Console links and connect your Search Console property.

You'll get two reports inside GA4: Google Organic Search Queries and Google Organic Search Traffic. The second one shows landing pages with Search Console metrics and GA4 engagement side by side. (Google Analytics Help)

Why does this matter for GEO?

Because AI Overview and AI Mode clicks are counted in Search Console's normal Performance report, and they end up in GA4's Organic Search channel. You can't separate them.

But you can take the top pages from Search Console's Generative AI report and look them up in the Google Organic Search Traffic report. That tells you how visitors to your most AI-visible pages actually behave.

A few limits to know:

• One GA4 web data stream can link to one Search Console property. (Google Analytics Help)
• Search Console data arrives in GA4 about 48 hours after collection. (Google Analytics Help)
• There's no specific AI Overview or AI Mode data in this integration.

Part 3: Fill the Gaps

GA4 will never see all your AI traffic. These three techniques help you estimate the rest.

#9: Ask people (“How did you hear about us?”)

This is the oldest trick in B2B marketing. And it's more useful now than ever.

Add a “How did you hear about us?” field to your demo and contact forms, with options like:

• Google search
• ChatGPT, Gemini or another AI assistant
• LinkedIn
• Referral from a colleague
• Event
• Other

Then send the answer to GA4 as an event parameter on your form submission event, and register it as a custom dimension (Admin → Data display → Custom definitions).

Now you can see AI-influenced leads even when the referrer was lost, and even when the buyer clicked a Google result after seeing your name in ChatGPT.

For many B2B companies, this is the single most useful AI attribution signal available. Because B2B journeys are long, and very few buyers click straight from an AI answer to a demo form.

#10: Track text-fragment clicks as a proxy for AI Overviews

This is a more technical one. But it's clever.

When Google links to a specific passage on your page, it often adds a text fragment to the URL. It looks like this:

https://example.com/page#:~:text=the%20exact%20passage

Links in AI Overviews often use these fragments.

GA4 doesn't record the fragment by default. But you can capture it with Google Tag Manager: read it with a bit of JavaScript, then send a GA4 event with the highlighted text as a parameter.

Be honest about what this is. It's a proxy, not a measurement:

• The person who published the method found text fragments in around 50% of cases. (Malcolm Gibb)
• Featured snippets and other Google features can use text fragments too, so not every fragment click comes from an AI Overview.

Still, even partial data is useful. The captured text shows you which exact passages Google is quoting from your page. That tells you what kind of writing gets cited.

#11: Watch Direct traffic to deep pages

When an AI tool strips the referrer, the visit usually shows up as Direct.

Direct traffic to your homepage is normal. People type your URL.

But Direct traffic to a deep page, like a 3,000-word guide or a specific comparison page, is much harder to explain. People don't type those URLs from memory.

So:

1. Go to Reports → Engagement → Landing page.
2. Filter to the Direct channel.
3. Exclude your homepage and main navigation pages.
4. Look for deep content pages with Direct traffic that's growing over time.

If those same pages are also getting AI Search traffic (Technique #5) and AI impressions in Search Console, there's a good chance some of that Direct traffic is AI traffic without a referrer.

Important: This is an educated guess, not proof. Internal email links, Slack and Teams shares and bookmarks can also create Direct traffic to deep pages. Use it as a directional signal alongside the other techniques, not as a number to report on its own.

Your GA4 GEO Checklist

Here's everything in one place:

Final Thoughts

GA4 won't give you a perfect picture of AI search. No tool does yet.

But put together, the AI Assistant channel, a well-built custom channel group, self-reported attribution and a few smart proxies give you something most B2B companies don't have: a realistic view of what AI visibility is actually worth.

Combine it with Search Console's Generative AI report and a regular citation audit in Google, and you've got a complete GEO measurement setup that doesn't cost a penny in extra tools.

If you'd like help setting it up, take a look at our AI Search Audits.

← Back to all posts