How to Track LLM Traffic in GA4 (And What You're Missing Without It)

fuse-smo-martin-janecekWritten by Martin J.
Back to blog
GA4 LLM traffic channel groups dashboard — ChatGPT, Perplexity, Claude sessions, dark UI

A significant portion of your GA4 "direct / (none)" traffic isn't direct at all — it's ChatGPT, Perplexity, and Claude, quietly misclassified. The question isn't whether LLM traffic is reaching your site — it almost certainly is. Are you measuring it?

Your GA4 dashboard was built for a world where users clicked links in search results or typed URLs directly. That world hasn't gone away — but something significant has been added to it. AI assistants now answer questions by recommending specific tools, articles, and companies. When a user follows that recommendation, they land on your site. GA4 records it as direct traffic. You record it as nothing. And somewhere in your strategy document, you're probably underinvesting in the one channel that's quietly sending you warm, high-intent visitors from people who were literally told to visit you.

That's the uncomfortable part. It's not that GA4 is broken. It's that GA4 was never designed to handle this, and nobody told you.

Here's how to fix it — and what you're likely missing right now.

What Is LLM Traffic and Why GA4 Doesn't Show It By Default

LLM traffic is any visit to your site that originates from an AI language model — ChatGPT, Perplexity, Claude, Gemini, Copilot — where the model cited or recommended your page in a response. The user then clicked through to you.

The attribution problem is technical, not mysterious. Most LLM interfaces don't pass a referrer header when a user navigates from an AI-generated response to an external site. Some strip it for privacy reasons. Others simply don't implement it. The result: GA4 receives a session with no referrer, no UTM parameters, and no campaign signal. It classifies it as direct / (none).

According to Orbit Media's 2025 traffic analysis, an estimated 25–40% of what currently appears as "direct" traffic in many content-heavy sites may actually be AI-referred visits. SimilarWeb's Q1 2026 data shows ChatGPT alone now accounts for approximately 1.2% of total referral traffic in certain content and SaaS niches — a figure that doesn't appear in most GA4 dashboards because it's buried under "direct."

This matters for three reasons:

  1. Budget allocation — If you can't see LLM traffic, you can't justify investing in AI visibility content
  2. Content strategy — LLM-referred visitors often have different intent profiles than search visitors; they were pre-qualified by the AI
  3. Competitive intelligence — Knowing which of your pages gets cited by AI assistants tells you what your strongest authority signals actually are

Understanding the "Direct / (None)" LLM Traffic Problem

Before you can fix the attribution, you need to understand exactly where it breaks.

When a user opens ChatGPT, asks a question, receives a response that includes a link to your article, and clicks it — the browser sends an HTTP request to your server. Whether it includes a Referer header depends on the browser, the LLM interface, and whether the link was clicked in a web app, a mobile app, or a copied URL.

In practice:

  • ChatGPT web interface: Referrer header is inconsistently passed. Many sessions land as direct / (none) in GA4
  • Perplexity.ai: Has been observed sending perplexity.ai as referrer in some sessions — but not all
  • Claude.ai: Typically strips referrer headers, almost always appearing as direct
  • Gemini: Behavior varies by interface (web vs. Google Search integration vs. app)

The result is a "dark traffic" pool that's genuinely difficult to disaggregate from true direct traffic (bookmarks, typed URLs, email clients that strip referrers).

Your best signal for identifying LLM traffic in GA4 today is a combination of:

  • Session source/medium containing known AI domains (chat.openai.com, perplexity.ai, claude.ai)
  • Unusual spikes in direct traffic correlated with content that ranks for AI-cited topics
  • Page-level direct traffic ratios that don't match the page's historical pattern

How to Track AI Traffic in GA4: Step-by-Step Setup

Here's the practical setup to capture whatever LLM referral traffic is passing a referrer header, and to create a reporting structure that separates it from true direct.

Step 1: Create a Custom Channel Group for AI Sources

In GA4, go to Admin → Data Display → Channel Groups → Create new channel group.

Add a channel called "AI / LLM Referral" with the following rules (use OR logic between sources):

Condition

Value

Session source contains

openai.com

Session source contains

chat.openai.com

Session source contains

perplexity.ai

Session source contains

claude.ai

Session source contains

bard.google.com

Session source contains

gemini.google.com

Session source contains

copilot.microsoft.com

Session source contains

you.com

Save this channel group. It will retroactively apply to historical data within your GA4 property window.

GA4 Custom Channel Groups configuration — LLM sources regex setup, dark dashboard

Step 2: Build an Exploration Report

Navigate to Explore → Blank exploration. Configure:

  • Dimensions: Session source / medium, Landing page, Channel grouping
  • Metrics: Sessions, Engaged sessions, Engagement rate, Conversions
  • Filter: Apply your "AI / LLM Referral" channel group, OR set a filter for session source containing the AI domains above

This gives you a direct view of which pages are receiving LLM-attributed traffic and how those sessions behave compared to other channels.

Step 3: UTM-Tag Your AI Citations (Where Possible)

For content you actively promote through your own AI-cited profiles or that you know gets picked up in AI responses — create UTM-tagged landing page variants or use canonical UTMs in your internal linking:

```
?utm_source=chatgpt&utm_medium=ai-referral&utm_campaign=llm-attribution
```

Some tools (including certain prompt-injection strategies for AI training data attribution) allow you to embed citation-friendly UTM URLs in structured data or sitemaps — though these are emerging techniques with inconsistent support.

Step 4: Monitor Direct Traffic Anomalies

Set up a GA4 alert for unusual spikes in direct / (none) traffic to specific pages. When a new LLM AI model starts citing your content at scale, you'll often see a sharp, page-level spike in direct traffic that doesn't match search or social patterns. This is your early-warning system for untracked LLM citations.

ChatGPT Referral Traffic: What It Looks Like in GA4

When ChatGPT does pass a referrer — which happens more consistently in the ChatGPT web app with certain browser/extension combinations — your GA4 will show session source as chat.openai.com.

Here's what to look for in your Acquisition → Traffic acquisition report:

  • Source/medium: chat.openai.com / referral
  • Default channel group: Often miscategorized as "Referral" rather than a distinct AI channel — which is why your custom channel group matters
  • Behavior signal: LLM-referred sessions from ChatGPT tend to have higher engagement rates (typically 15–25% above site average in SaaS content) because users arrived with specific intent, pre-qualified by the AI's recommendation

Perplexity traffic, when attributed, shows as perplexity.ai / referral. It's more reliably tracked than ChatGPT because Perplexity's web interface passes referrer headers more consistently. If you're seeing Perplexity traffic, your GA4 is actually showing you a cleaner slice of your LLM attribution picture than most sites have access to.

LLM referral traffic analytics — ChatGPT vs Perplexity vs Claude comparison chart, dark UI

The honest assessment: even with this setup, you're likely seeing 20–60% of your actual LLM traffic depending on your audience's device mix, browser settings, and which AI tools they use. The rest stays in direct / (none). For a deeper look at how ChatGPT-specific citation patterns behave, see our ChatGPT visibility tracking article.

Beyond GA4: How Allable.ai Automates LLM Traffic Attribution

The manual GA4 setup above gets you a partial picture. It requires ongoing maintenance as new AI sources emerge, and it fundamentally can't recover the traffic that never passed a referrer header.

Allable.ai's AI Visibility feature approaches this differently. Rather than trying to reconstruct attribution from broken referrer signals, it monitors where your content is being cited within AI responses directly — tracking which of your pages appear in ChatGPT, Perplexity, Claude, and Gemini outputs for relevant queries, and at what frequency.

This gives you two things GA4 can't:

  1. Proactive citation tracking — You know which pages AI assistants are citing before users even visit your site, so you can double down on content that's already building AI authority
  2. Gap identification — You can see which high-intent queries in your space are being answered by AI assistants, whether your content is appearing, and specifically what's missing that would make AI more likely to cite you

The attribution picture becomes: GA4 for the traffic you can measure, Allable for the citation signal that precedes and predicts that traffic.

Pricing: Free forever plan available; Pro at $33/month, Business at $98/month — with AI visibility monitoring included from the Pro tier.

For context on the full AI visibility opportunity, our LLM tracking tools guide walks through how citation frequency translates to organic visibility across AI-powered search surfaces.

Frequently Asked Questions

How do I see ChatGPT traffic in GA4?
In GA4, go to Reports → Acquisition → Traffic acquisition and filter session source by chat.openai.com. If you're seeing sessions, ChatGPT's web interface passed a referrer header for those visits. For a complete picture, create a custom channel group that includes all major AI source domains — most ChatGPT traffic that doesn't pass a referrer will still appear under direct / (none), which you'll need to infer from traffic spike analysis.
Why does GA4 show LLM traffic as direct?
GA4 records direct traffic when there is no referrer header in the session. Most LLM interfaces — particularly Claude, ChatGPT mobile, and embedded AI assistants — strip the referrer for privacy or technical reasons. The result is that a session initiated from an AI recommendation lands in GA4 with no source, no medium, and no campaign — indistinguishable from a typed URL or bookmarked visit.
Is all my direct traffic potentially LLM traffic?
Not all of it — but more than most teams account for. True direct traffic (typed URLs, bookmarks, email clients) still exists and is often the dominant component. The signal to watch for is page-level anomalies: if a specific article you haven't promoted elsewhere suddenly spikes in direct traffic, especially if that content covers a topic AI assistants frequently discuss, LLM referral is a probable contributor.
Can I track Perplexity traffic separately in GA4?
Yes — Perplexity more consistently passes a referrer header than ChatGPT. In GA4 Traffic acquisition, filter for session source containing perplexity.ai. You can also include it in your custom AI/LLM channel group to aggregate it with other AI sources in a single reporting view.

What to Do Next

The GA4 setup in this guide gives you the best available visibility into LLM traffic within your current analytics stack. It won't capture everything — that's a structural limitation of referrer-based attribution in AI interfaces. But it gives you a working baseline, a custom channel group you can build on, and the anomaly monitoring to catch new LLM citation waves as they arrive.

Your competitors are already using AllAble. Are you?

The marketers pulling ahead aren't working harder. They're just working with one tool that does everything — that tool is AllAble. Try it yourself!