When a user clicks a citation link inside an AI assistant, tracking that visitor requires understanding how different LLM traffic sources in GA4 report their data. You can isolate these conversational clicks by deploying regex filters, building custom exploration reports, or configuring a dedicated custom channel group. Because major AI assistants do not share a single referral standard, combining these custom setups with active traffic generation is the only way to reliably measure and grow your conversational AI referrals.
To track AI referrals accurately, you must set up a custom channel group in GA4 using source rules that match specific assistant referrers like chatgpt, claude, gemini, and perplexity. This prevents conversational search clicks from being misclassified as generic direct or referral traffic.
What Counts as LLM Traffic in GA4
LLM traffic is the volume of visitors who arrive at your website by clicking a link, citation, or recommendation embedded within an AI assistant's chat response.
This flow of visitors comes from engines like ChatGPT, Gemini, Claude, and Perplexity. When a user asks an AI assistant for a tool recommendation, a local service, or a source, the AI often provides a clickable link. If the user clicks that link and lands on your site, GA4 registers this session. This specific user action is what you want to isolate when identifying LLM traffic inside your GA4 reports.
Crucially, this is distinct from mere AI visibility. Having your brand cited or mentioned in an LLM response is excellent for brand awareness, but it does not affect your analytics until a user actually clicks through.
How GA4 Currently Classifies This Traffic
Google Analytics 4 does not feature a pre-built default channel group specifically for AI assistants. Visits from these platforms get scattered across different default categories like Referral or Direct depending on how the assistant passes user data. Because source and medium values differ dramatically from one assistant to the next, relying on default reports will fragment your data.
When an assistant does pass a referrer, GA4 can label the source clearly enough to identify and analyze the users originating from chats, as noted in an analysis of LLM traffic by Wix Studio. To capture everything accurately, most marketing teams avoid default channels and focus on setting up AI referral traffic tracking in GA4 from scratch.
Where Each Major AI Assistant Tends to Show Up in GA4
AI assistants do not share a single, standardized method for passing traffic data. Some platforms preserve the referrer domain when a user clicks an outbound link, allowing Google Analytics 4 to easily group them under referral traffic. Others strip this data entirely.
A single catch-all filter will miss a significant portion of your AI-driven visitors. Understanding why ChatGPT sessions often land in the direct traffic bucket is the first step toward correcting these attribution gaps. You will face a similar challenge when confirming Claude visits that GA4 is misreading as direct.
Google's own ecosystem introduces further complexity. Marketers must focus on separating Gemini traffic from Google organic search to prevent conversational AI clicks from inflating traditional search engine metrics.
| AI Assistant | Referrer Header Status | Primary GA4 Channel Allocation |
|---|---|---|
| ChatGPT | Inconsistent depending on app vs. web usage | Referral or Direct |
| Claude | Often stripped completely | Direct |
| Gemini | Varies based on integration type | Organic Search or Referral |
| Perplexity | Generally preserved | Referral |
Three Ways to Isolate LLM Traffic Sources in GA4
Regex filter in standard reports
To get a quick overview of your conversational AI traffic without altering your permanent GA4 configuration, you can apply a regular expression filter directly inside your standard Traffic Acquisition reports. This method isolates traffic using known referrer strings like chatgpt, claude, gemini, or perplexity within the Session source/medium dimension.
Exploration report for deeper analysis
Standard reports limit your ability to cross-reference AI traffic with specific landing pages or user behavior. Creating a blank Exploration report gives you full control over dimensions, metrics, and segments. This method lets you build a dedicated workspace where you can analyze conversion rates and user journeys specifically for visitors coming from LLM sources.
Custom channel group for a permanent view
For a permanent solution that integrates directly into your default reporting, you can define a custom channel group. An agency strategist shared on Reddit how they added LLM traffic as a custom channel category to all of their client sites in GA4. Doing this ensures the data is retroactively processed and persistently visible.
To set this up, follow the instructions for building the exploration report and custom channel group step by step:
- Navigate to your GA4 Admin panel and locate the Data Settings menu.
- Select Channel Groups to define a new custom group or edit your existing one.
- Create a new channel rule named "AI Referrals" or "LLM Traffic".
- Define the rule conditions using source names matching your target AI assistants.
- Save the channel group and apply it as your primary reporting dimension.
What These Tracking Methods Still Miss
Implementations of custom channel groups and regex filters only capture the traffic that actually leaves a digital paper trail. If an AI assistant strips the referrer header entirely—which occurs regularly with mobile apps—the visit drops directly into your unclassified direct bucket.
Your tracking rules are also limited by what you manually define. When a new conversational assistant launches or an existing one changes its referral domain format, those clicks will slip past your filters unnoticed until you manually update your settings.
Acknowledging these blind spots is essential when comparing LLM referral volume against traditional search traffic. Because conversational AI attribution is naturally leaky, your analytics will almost always underreport the true impact of LLM referrals.
Turning LLM Traffic Sources Into a Channel You Can Grow
Once you isolate LLM traffic in your reports, you can treat it as an active acquisition channel to grow rather than a passive metric to monitor. PerkFuel enables this shift by sending targeted visits from AI assistants to specific pages on your site. You can choose from 12 AI sources, including ChatGPT, Gemini, Claude, Perplexity, Copilot, Grok, Meta AI, DeepSeek, Mistral, Manus, Poe, and Kimi.
This setup allows you to target visitors from 65+ countries and control the pace of delivery. Under the Launch plan, which costs $49/month, you receive 300 visits, with support for up to 3 pages per project, up to态 projects, and up to 30 visits/day per project. To test the service, new accounts receive 10 free visits with no credit card required.
All campaigns deliver real visitors only. These visits are tracked inside the PerkFuel dashboard and, once you enable analytics tracking, they will appear in Google Analytics as well. While PerkFuel guarantees the delivery of the visits included in your selected plan, it does not guarantee rankings, engagement, or conversions. Those outcomes depend on your specific website and offer.
Configure your custom channel groups in Google Analytics today to stop conversational search clicks from masking themselves as anonymous direct traffic. Remember that mobile apps and stripped referrer headers will always cause some tracking leakage. Once your tracking framework is in place, you can actively scale this channel by using PerkFuel to deliver real, targeted AI referral visits directly to your highest-value landing pages.