GA4 shows AI traffic, but can't tell you why

To most marketers, GA4 plays an indispensable role in their marketing stack. But while GA4 remains a powerful tool in the right hands, relying solely on it to capture AI search data comes with several challenges. This is because GA4 cannot show you what prompt generated that visit, how your brand was described in AI answers, or whether a competitor was recommended more favourably in the same response. These are structural limitations that no custom setup can resolve.

For teams that want to understand their brand’s AI search visibility, this creates two challenges: getting GA4 configured correctly so the traffic you can measure is measured meaningfully, and recognising where GA4's usefulness in AI visibility monitoring ends.

In this article, we cover what GA4 can reliably track for AI referral traffic, the proxy signals available in your existing tools, and what a purpose-built AI visibility system adds beyond them.

What GA4 can and cannot track for AI referral traffic

Let’s address the elephant in the room first: GA4 does not automatically separate AI referral traffic from regular referrals. Without a custom channel group configured, ChatGPT sessions typically appear inside the Referral channel, occasionally under Direct, or as chatgpt.com / (not set) in the unassigned bucket. This happens because ChatGPT sometimes passes a utm_source parameter without a corresponding utm_medium, which GA4 cannot cleanly resolve.

Often, the recommended fix is setting up a custom channel group: a regex filter applied in Admin that captures known AI domains and assigns them to a dedicated AI channel. The channel should be positioned above Referral in the evaluation order, since GA4 processes rules from the top down and will otherwise classify AI traffic as a generic referral first. Setup takes around fifteen minutes and applies retroactively to historical data, which means you don’t need to wait for new traffic to start seeing patterns.

Once that configuration is in place, GA4 can tell you:

●     Which AI platforms are sending traffic, and the volume over time

●     Which landing pages those sessions arrive on

●     How AI-referred sessions behave: engagement rate, time on site, pages per session

●     Whether AI-referred sessions appear in multi-touch conversion paths

That’s, without a doubt, genuinely useful information. AI-referred visitors consistently show stronger engagement signals than average organic traffic, with lower bounce rates and longer session durations. If AI referral volume is growing and those sessions are appearing in conversion paths, that’s a meaningful signal worth tracking.

There are, however, four structural gaps that no configuration change on GA4 can close.

No query data. GA4 captures the referring domain, not the prompt that generated the visit. The query that sent someone to your site is unavailable in any standard or custom GA4 report.

A significant share of AI traffic lands as Direct. Mobile app users on ChatGPT and Claude frequently strip referrer headers before the session reaches GA4. GA4's client-side JavaScript also can’t detect AI bot crawls that access your content without generating a browser visit. Industry analysis suggests that true AI influence on traffic is likely two to three times what analytics reports show, because these indirect pathways aren’t attributed.

Google AI Overviews are invisible in GA4. Clicks from AI Overviews and Google AI Mode register as google / organic, indistinguishable from standard organic search traffic. Google added an AI Mode filter to Search Console in June 2025, but this shows impressions within Google's AI responses rather than feeding into GA4 as a separate source. At the time of writing, there’s no way to isolate this traffic in standard reporting.

Misclassification remains the default without configuration. Even with a custom channel group in place, sessions from the ChatGPT mobile app or AI tools that suppress referrer data will still land as Direct. Some AI platforms also use internal redirects or proxy servers that intentionally strip referrer data for privacy reasons.

In essence, GA4 can't tell you what your brand looks like in the answer that sent it, whether you’re winning or losing share of voice against competitors, or whether you’re being recommended as a preferred solution or mentioned in passing. Those require a different kind of setup to meaningfully track. 

Why AI referral traffic often lands on your homepage

When teams first examine their AI referral landing pages in GA4, one observation they make is that a disproportionate share of sessions lands on the homepage rather than specific pages that should logically be getting cited.

There are a few reasons this happens. 

AI answers frequently mention a brand by name without including a link to a specific URL. A user sees the recommendation, doesn’t click a citation link, and navigates directly to the homepage. That session registers as Direct in GA4, not as an AI referral. When AI platforms do pass a referral, they sometimes link to the root domain, particularly in recommendations where the AI answer names the brand without citing a specific piece of content.

Branded searches that follow an AI mention also follow a similar pattern. Someone sees your brand in a ChatGPT answer, doesn’t click, and instead searches for your brand name on Google moments later. GA4 attributes that session to organic, even though an AI answer initiated the consideration.

This means that homepage landing rates and Direct traffic trends are both indicators of AI influence that aren’t captured by referral attribution. What looks like a modest AI referral channel in GA4 is often part of a larger pattern of AI-influenced discovery happening just outside of what’s getting tracked. 

Building an AI search visibility map from buyer intent

On the second limitation, GA4 doesn’t show you where you stand across the full range of questions your buyers are actually asking in answer engines. Understanding that requires starting from your buyer’s intent and working backwards to the questions they’re likely asking AI.

Most buyers shift through three types of questions when making decisions. Mapping your AI visibility across all three tells you something qualitatively different from anything GA4 can show.

In practice, a brand that appears consistently in educational prompts but disappears in transactional or commercial ones likely has an intent gap, while a brand that is misdescribed likely has a positioning gap. These are distinct problems with different fixes, and GA4 data can’t reveal them in a way that’s actionable.

The landing pages already appearing in your GA4 AI referral report are a useful starting point for this exercise. They tell you which content the answer layer currently treats as relevant. Working backwards from those pages, you can identify what intent they’re serving and which related questions in the same topic cluster you’re not yet appearing for. That turns your existing analytics data into a broader visibility map rather than just a search traffic report.

For the framework to track the four core metrics across that map, including visibility rate, share of voice, sentiment distribution, and position-adjusted visibility, our piece on measuring brand presence in AI search covers the methodology in detail.

Proxy signals for AI search visibility in your existing tools

For teams not yet running systematic AI visibility tracking, there’re signals available in existing tools that approximate what’s happening. None of these are substitutes for direct measurement, but they, when used in tandem, can inform decision-making. 

Landing page clusters in GA4. Look at which pages are receiving AI referral traffic and ask whether they cluster around a particular topic, funnel stage, or content format. If comparison and alternatives pages appear consistently but problem-aware content does not, that’s a signal about where in the buyer journey AI’s currently citing you and where it’s not.

Branded search volume in Search Console. A rising trend in branded searches often follows increased AI visibility. Customers see your brand in an AI answer, don’t click through, and search for you directly later. Branded search trends are a lagging indicator of AI mention rate and tend to move in correlation with visibility improvements.

Assisted conversions in GA4. Examining how AI-referred sessions appear in multi-touch conversion paths reveals whether these sessions are functioning as awareness touchpoints earlier in the customer journey rather than last-click converters. Case in point: a session from Perplexity three days before a direct conversion represents AI-influenced consideration that last-click attribution will not capture. GA4's path exploration and attribution reports highlight these patterns. 

Unexplained lifts in Direct traffic. Correlated spikes in direct traffic following content updates, PR activity, or periods of increased category conversation can be a rough proxy for the brand-mention-driven navigation described in the previous section. This is naturally imprecise, but it’s a signal worth monitoring alongside the others.

The catch with these signals, of course, is that they indicate directional trends, not actual mention rates, competitive share of voice, sentiment, or a platform-by-platform breakdown of where you are visible versus absent. They are useful for identifying which content is already resonating with answer engines.

What a purpose-built AI visibility platform adds

Having a well-configured GA4 setup can provide valuable AI visibility insights, but those insights stop short of revealing more critical information: what your brand looks like inside the AI responses that are shaping buyer decisions before they ever arrive on your site. 

Specifically, GA4 cannot tell you:

●     Which prompts are generating your citations, and which ones your competitors are winning instead

●     Whether you are gaining or losing share of voice in your category over time

●     How your brand is being characterised in AI answers, whether accurately, generically, or with caveats

●     Which platforms are citing you versus not citing you at all

●     Whether content changes you have shipped have influenced any of these metrics

Resolving these issues require a more direct measurement of AI answers, and a means to follow up with these findings. At Wordflow, we connect AI visibility analysis directly to content engineering inside a single platform, so users easily identify fixes and create actionable content that closes the gap from weeks to hours. Go from insights to content that makes your brand show up in AI answers today.  

‍