How to Use GA4 to Spot AI-Driven Traffic

AI-powered search isn’t just another referral source. It behaves differently: it parasitically repackages content, is sometimes a little foggy about the attribution it gives, and will give markedly different answers to the same query, depending on circumstance. AI website traffic is, indeed, a different animal than traditional traffic.

In this post, we’re going to explore an essential set of exercises marketers should think about, as they seek to understand where AI-driven traffic is coming from. In the process of looking at this traffic, we’ll also uncover how that traffic represents the “AI cognitive bias” influencing which urls and content that the platforms tend to favor.

What follows is a blueprint for a proactive monitoring framework for Google Analytics to better understand AI traffic and, potentially, to influence which content on your site gets surfaced. The following exercises are designed to give you visibility, context, and evidence.

By the way: if you’d like to discuss how these apply to your own site, you can book a time to chat with me (for free) here.

1. Why AI search traffic is hard to see in GA4

Google Analytics was not originally designed to track traffic from AI platforms. Namely:

• Traditional GA4 channels are built around obvious referrers (Google, Bing, Facebook).
• LLM-driven referrals often come through odd hosts, redirects, or appear as “Direct.”

The result of all this is phantom traffic: real visits whose origin is seemingly cloaked. To solve this problem, we can create custom segments which track incoming traffic from AI platforms.

The exercise: Create a living GA4 segment to capture traffic from known AI domains.
I don’t see the point in writing a post showing you how to create custom segments in GA4 to track AI-driven referrals. Several others have already written about using regex strings to do so.

Here’s a post which offers a quick recipe: Two Octobers “Tracking AI Traffic in GA4.”

2. Build a Living “AI Referrers” Segment

Once you’ve identified phantom traffic (step 1), the next challenge is keeping that identification fresh as new AI players enter the field (step 2).

Identifying AI traffic once isn’t the real challenge — keeping track of it as platforms shrink and grow is. A one-time regex filter will need to be updated regularly, in order to have evergreen value.

Treat your segment as a living asset inside GA4, by doing the following:

• Review it monthly and update as new AI domains surface.
• Use flexible matching, not rigid lists. GA4 lets you write patterns that match variations of domains instead of typing every possible one out. For example, instead of entering “chat.openai.com” and “chatgpt.com” separately, you can write a single pattern that captures both. This means you won’t need to rebuild your filter every time a new version of the domain shows up.
• Save and share the segment so it’s reusable. Once you’ve built your AI traffic segment, GA4 allows you to save it and make it available in multiple reports. That way you don’t have to recreate the same filter each time, and your colleagues can use it too. It’s like having one master switch that everyone can flip on when they want to see AI-driven traffic.
• If you haven’t used this feature before, this guide to creating and managing segments in GA4 walks through the steps of saving a segment so it can be applied across reports and shared with colleagues.

Why it matters:
By institutionalizing the segment, you turn what could be a fragile, one-off filter into a repeatable framework. Your system has the built-in capacity to evolve with the ecosystem of search.

3. Compare AI vs. Human Referrals by Landing Page

There is often a mismatch between the pages marketers and brands would like a new visitor to encounter and where AI will often lead them. AI tends to surface content which answers questions, whereas humans searching in Google often arrive with the intent to evaluate services.

• Platforms like Perplexity and ChatGPT tend to link to “answer-style” content — FAQs, glossaries, quick guides — not your practice or service pages. (This preference was the reason behind a recent post of ours.)
• A landing page report split by AI vs. traditional referrals shows what content AI is surfacing.
• The mismatch: AI may be amplifying low-value pages that weren’t meant to convert.

Why it matters:
A landing page is simply the first page a visitor sees. In a perfect world, that’s your service overview, a lead form, or a key article. But if AI is consistently driving people to minor blog posts or definitional content, it changes how you think about funneling those visitors. This reality creates a challenge for marketers in that successful AI-forward strategies might yield more traffic. But that traffic will now be entering the back door. This might govern a shift in strategy: how do we redirect the user journey from the pages AI loves, to the pages we want new visitors, prospects, etc. to see?

The exercise: Segment AI traffic and run a landing page report. Compare which pages AI visitors land on versus traditional referrals. If the AI pattern is different, decide whether to beef up those pages, redesign their calls-to-action, or adjust your strategy so the traffic doesn’t go to waste.

4. Detect AI Disguised as Direct Traffic

The fog of attribution persists in other ways. AI referrals often appear as “Direct” in GA4, with no clear referrer. To spot some of these:

• Look for spikes in Direct sessions with odd characteristics: short visits, high bounce, unusual entry points.
• Compare trendlines before and after major AI features roll out.

Why it matters: Direct traffic is a junk drawer. Some of it is genuine (people typing your URL), but a surprising share is just traffic GA4 can’t figure out the source of. AI fits into that second category. If you’ve ever seen a spike of Direct visits landing deep inside your site — pages no one would type in manually — that’s a sign AI may be behind it.

As a marketer, if you start seeing unlikely pages—such as an innocuous FAQ page— suddenly showing spikes in Direct traffic, look for correlations to what’s happening in AI search. These are the sorts of events which are typical after, say, a new feature rollout from Google or one of the other AI platforms.

The exercise: Track your Direct traffic over time. Flag any sudden increases, especially after high-profile AI announcements (e.g., Google AI Overviews expanding to new markets). Treat these unexplained surges as a proxy for AI traffic until analytics tools catch up.

5. Custom Events for “AI-like” Visits

Another way to get a handle on AI-driven visits is to create custom events which identify and track behaviors which represent incoming AI traffic. For instance:

• Some AI-driven visits won’t carry a clear label. When a visitor clicks through from Google, Facebook, or Bing, GA4 usually records the source domain, so you can tell where they came from. But when a click originates inside an AI platform, the referral information often gets stripped away. That means you don’t see “chatgpt.com” or “perplexity.ai” in your reports — instead, the visit may look like it came from “Direct” or from a meaningless string of characters. In practice, this makes AI traffic invisible unless you deliberately set up ways to flag these sessions.
• Some referrals look strange or incomplete. In a typical analytics report, referrer strings follow a predictable format (like google.com). With AI traffic, you may see half-formed domains, odd subdomains, or even fragments of an app’s internal structure. These are signals that the visit didn’t come through traditional search or social.
• Visitors may land deep in your site without context. If someone types your firm’s name into Google, they’ll usually land on your homepage or a main service page. But AI visitors often appear suddenly on obscure internal pages — the kind no one would type in manually. This “teleporting” behavior is a clue that an AI engine dropped them there directly from a generated answer.

Why it matters: Events in GA4 are often used to track conversion activities such as clicks and form fills. They can be used more creatively, though, and here they can be used as a way to flag traffic anomalies. By turning suspicious session patterns into custom events, you create a breadcrumb trail — a dataset of likely AI referrals even when the source is hidden.

The exercise: In GA4 (or better yet, through Google Tag Manager), create events that fire when visits meet “AI-like” criteria: missing attribution, strange referrers, or deep-page landings. Over time, you’ll build a pool of traffic that behaves differently, which you can then compare against other sources.

A really solid article illustrating how this can be done is here.

6. Cohort Analysis of AI Visitors

Understanding the volume of AI-driven traffic is useful, but the real question is whether those visitors create business value. GA4’s cohort analysis feature allows you to group visitors by source and compare how they behave over time.

• A cohort is simply a group of visitors who share a characteristic (such as their traffic source or the week they arrived).
• Cohorts let you see how different groups perform side by side, showing engagement patterns and conversion trends.
• For AI traffic, this can answer the critical question: do these visitors engage meaningfully, or do they disappear after a single visit?

Why it matters: Executives and partners want to know not just “how much” but “how valuable.” Cohort analysis provides evidence: are AI visitors signing up, contacting you, or returning — or are they just one-time readers?

The exercise: In GA4 Explorations, build two cohorts: one for visitors arriving via your AI-referral segment and one for organic search visitors. Compare metrics such as conversion rate, engagement time, and return visits. This shows whether AI-driven traffic is worth optimizing for.

Further reading: OptimizeSmart – Cohort Exploration Report in GA4

7. Monitor AI Search as Its Own Channel

GA4 lumps AI traffic into categories like Direct or Referral, making it invisible at the reporting level where most stakeholders pay attention. By treating AI as its own channel, you give it equal standing with Organic, Paid, and Social.

• Executives interpret performance in terms of channels, not regex filters or referrer strings.
• Custom channel groupings in GA4 let you define a new channel, such as “AI Search.”
• Once created, AI traffic shows up as a distinct line item in all your channel reports.

Why it matters: When you present AI traffic as a proper channel, it stops being a technical footnote and becomes a strategic metric. Saying “AI Search contributed 4% of sessions this month” communicates clearly to leadership and clients.

The exercise:In GA4, go to Admin → Data Settings → Channel Groups. Create a new channel group that buckets your AI referrer segment into a channel called “AI Search.” From then on, AI will appear alongside your other core channels.

Further reading: MarTech – Why it’s time to treat AI referrals as their own channel in GA4

Exit Strategy

AI search traffic is messy, inconsistent, and often invisible. But with a systematic approach, you can bring it into view. Most of us work in environments where the analyses we’re asked to bring to stakeholders are critical for performance evaluation and changes to marketing strategy. In the case of something as transformative as AI, the volatility of the current search ecosystem demands creative methods to track influence and visitation.

In short: AI traffic is not an anomaly; it’s a growing channel. If you don’t track it now, you’ll misattribute a bigger slice of future traffic.

If you’d like help building this framework in your own analytics, I’d be glad to walk through it with you: https://calendly.com/tyrannoceratops.

FAQ: Tracking AI Traffic in GA4

1. Can GA4 automatically detect AI traffic for me?
Not yet. GA4 doesn’t have a built-in channel for AI search or chatbot referrals. Right now, the only way to see this traffic is to create your own segments, events, or channel groupings that surface the patterns AI leaves behind.
2. What’s the difference between AI traffic and traditional referral traffic?
Traditional referrals come with clear labels — for example, facebook.com / referral or bing.com / organic. AI referrals often strip that data away. As a result, they may show up as “Direct” traffic or with odd, partial referrer strings that don’t look like normal websites.
3. Why should my firm care about AI traffic if conversions are still low?
Even if AI visitors don’t convert today, they are an early signal of how search behavior is changing. Just as mobile traffic started small before becoming dominant, AI-driven clicks may represent the front edge of a much larger trend. Watching them now gives you a head start.
4. What if I don’t have the resources to set up custom events or cohorts?
Start small. Even creating a single segment to capture known AI domains (like chat.openai.com or perplexity.ai) can give you valuable visibility. As you get more comfortable, you can layer in the more advanced exercises described in this post.
5. Do I need to switch away from GA4 to track AI traffic better?
No. GA4 is still the best general analytics platform for most organizations, and with custom configuration it can handle AI traffic just fine. If you need more flexibility, you can extend GA4 with tools like Google Tag Manager or BigQuery — but switching platforms isn’t necessary.