Your AI referral source data looks active, yet the numbers are fundamentally flawed. This is a common pitfall in GA4 LLM tracking: teams apply the correct regex filters but pair the wrong dimension with the wrong metric, creating a logical contradiction that results in empty or misleading reports. The issue is not a lack of tracking methods; it is a failure to make one specific, high-impact architectural decision about how to define the origin of AI-driven visits.
The choice between Session Source and First User Source dictates whether you measure the volume of influenced sessions or the long-term value of acquired users. This article focuses on that critical distinction, ensuring your LLM analytics setup reflects reality rather than noise. By clarifying this foundational choice, you can move beyond superficial data collection to a strategic understanding of how generative search is reshaping your customer journey.
Session Source vs. First User Source: The Core Decision
In GA4 LLM tracking, the most common error is confusing two distinct attribution models that serve different purposes. Session Source, found in Traffic Acquisition, records the referrer for every individual visit. First User Source, found in User Acquisition, locks in the channel that initiated a user’s very first interaction with your domain. This distinction is critical for an accurate LLM analytics setup.
Consider a user who discovers your site through a recommendation in ChatGPT. That interaction is captured as the First User Source. If the same user returns two weeks later by clicking a Google Ad, the Session Source for that second visit is the ad network, not the LLM. The AI engine gets credit for the initial discovery, while the ad gets credit for the re-engagement.
You must align these dimensions with their corresponding metrics to avoid data integrity issues. The rule is absolute: pair Session Source with the metric Sessions, and First User Source with Total Users.
| Dimension | Report Location | Required Metric | Use Case |
|---|---|---|---|
| Session Source | Traffic Acquisition | Sessions | Volume of AI-driven visits |
| First User Source | User Acquisition | Total Users | Long-term AI user acquisition |
Mixing these—for example, applying Session Source to Total Users—creates a logical contradiction that GA4 cannot validate. This results in empty or misleading data, making it impossible to accurately track AI traffic in GA4. When defining your AI referral source, clarity in this pairing ensures you are measuring the right thing at the right level of granularity.
Applying the Regex to Your AI Referral Source Dimension
To begin tracking AI traffic in GA4, you need to filter out the noise from specific Large Language Model referrers. The standard regex pattern used for this setup contains 17 distinct rules to capture the major players in the AI space. It covers specific platform domains like ChatGPT, Gemini, Claude, and Copilot, while also catching generic variants through the .*\.ai$ rule.
For a proper LLM analytics setup, you should configure this directly within your standard reports rather than relying on ad-hoc explorations. This ensures that the data is consistently available for regular monitoring without having to rebuild the logic every time.
Configuring the Session Source Filter
The most common approach is to apply this filter to the Session Source dimension. This captures every individual visit that originates from an AI tool, regardless of whether the user was new or returning. To apply this, follow these specific steps within the GA4 interface:
- Navigate to the Acquisition section and select Traffic Acquisition.
- In the dimensions row, ensure Session Source is displayed.
- Click the plus icon next to the dimension to add a filter.
- Select Session Source as the dimension.
- Choose matches regex as the match type.
- Paste the 17-pattern regex string into the input field.
This configuration gives you a clear view of the total volume of traffic driven by AI tools in real-time. It is the most responsive signal for checking how your content performs in AI-generated answers.
The First User Source Alternative
While Session Source measures volume, the First User Source dimension offers a different perspective. You can apply the exact same regex pattern to this dimension, but you must do so within the User Acquisition report. This setup isolates the unique users whose very first interaction with your site was via an AI tool.
The key difference lies in interpretation. If a user comes in from ChatGPT today, Session Source counts it. If they return next week via Google, Session Source switches to Google, but First User Source keeps them attributed to the AI tool that originally brought them in. For many teams, the regex is identical, but the story it tells shifts from “daily traffic” to “long-term acquisition.”
Multi-Session Journeys and AI Traffic Attribution
The choice between these dimensions fundamentally shapes your attribution model. If your goal is to measure the total volume of visits influenced by AI, Session Source is the correct lens. It captures every instance where an LLM referred a user, regardless of when it happened in their history. Conversely, if you aim to measure the long-term value of users first acquired by AI, First User Source is the appropriate metric. This dimension locks in the initial touchpoint, ensuring that the AI platform receives credit even if the user later interacts with your site through other channels.
Consider a concrete scenario to see how this plays out. A user visits your site via ChatGPT in Session 1, then leaves and returns later via a Google Search result in Session 2. Under Session Source logic, the LLM gets credit for Session 1, while Google receives credit for Session 2. This split reflects the immediate context of each visit. However, under First User Source, the LLM retains the credit for the entire user journey because it was the initial point of contact. This distinction is critical when assessing the true origin of customer relationships versus immediate traffic spikes.
For most brands, tracking both metrics provides the most complete picture. You can monitor daily traffic fluctuations through session-based data while evaluating long-term acquisition health through user-based data. That said, if you need a quick health check to answer the question “is AI driving traffic?”, Session Source offers the most immediate and responsive signal. It directly correlates with recent content performance and current LLM recommendations, making it ideal for short-term tactical adjustments. First User Source is better suited for strategic planning and measuring the lasting impact of AI as a channel over time.
Common Questions on GA4 LLM Tracking Setup
A clear understanding of how LLMs interact with your website prevents common tracking errors. Here are answers to the most frequent questions regarding this GA4 LLM tracking configuration.
Do You Need UTM Parameters?
No. Unlike social media platforms, LLMs do not append UTM tags to outbound links. You cannot rely on utm_source or utm_medium to identify this traffic. Instead, the tracking logic must depend on the AI referral source domain (Source/Medium) and the regex filtering method described earlier. Attempting to force UTM parameters onto LLM links is ineffective because the AI controls the final URL structure.
Will Google Create a Native Channel?
Google has indicated plans for a dedicated AI traffic channel, but until that feature is fully released, the manual setup remains the standard. For precise, immediate data, the custom regex filter is the reliable approach. Waiting for a native feature means losing visibility on your current LLM analytics setup performance, so the manual method is currently the only way to get accurate numbers today.
How Often Should You Update the Regex?
Regularly. The AI landscape shifts quickly. As new models like Grok or DeepSeek gain market share, their domains change. You must review and update your pattern to ensure your dimension captures the full scope of emerging engines. A static list quickly becomes obsolete, leading to underreported AI-driven visits.
Strategic Takeaway
The right configuration for your GA4 LLM tracking depends entirely on the KPI you are trying to defend. If your priority is measuring immediate content performance or total visit volume, Session Source paired with Sessions offers the most responsive signal. If you are evaluating the long-term value of customer acquisition, First User Source paired with Total Users provides a more accurate picture of where new relationships begin.
There is no universally superior setting; the best choice is the one that aligns with your specific business objective. Remember that the regex will require periodic updates as new models emerge. However, the architectural decision between these two dimensions is foundational. It defines not just how you report, but how you ultimately perceive and value AI-driven growth in your strategy.
