How to Track LLM Traffic in GA4: Exploration Reports and Custom Channel Groups

Learning how to track LLM traffic in GA4 is essential to see which AI assistants are sending users to your landing pages. Isolate this traffic by building a custom exploration report with a regex filter, then make those insights permanent by setting up a custom channel group. This setup separates real human clicks from automated crawler bots, giving you clean data to measure your AI referral performance.

What counts as LLM traffic in GA4

LLM traffic in Google Analytics 4 (GA4) refers to user sessions initiated when a visitor clicks a link inside an AI assistant's response rather than a traditional search engine or social platform.

GA4 does not ship with a dedicated AI or LLM channel by default. These sessions land under generic referral categories or get absorbed elsewhere, obscuring how visitors actually arrive on a site through an AI assistant's answer.

The platforms sending this traffic extend far beyond the few most people think of. While ChatGPT, Gemini, Claude, and Perplexity are common, PerkFuel supports 12 AI sources including Copilot, Grok, Meta AI, DeepSeek, Mistral, Manus, Poe, and Kimi. To see these visits clearly, marketers must manually group them. Some analysts configure custom channel categories in GA4 specifically to isolate and track traffic across multiple LLM engines, as noted in the Google Analytics community on Reddit.

LLM referral traffic vs. LLM bot traffic

LLM referral traffic represents a human visitor who clicked a link generated by an AI assistant. These are genuine sessions associated with a physical device, a specific referrer, and active behavior on your site. LLM bot traffic consists of automated crawlers fetching your pages for indexing or model grounding.

Your custom channel groups and filters should target referral sessions specifically. Attempting to mix bot activity with referral data skews your engagement metrics, as bots do not interact with your tags or content the way humans do. Real AI referral visits may browse pages, scroll, and click links.

Track LLM traffic in GA4 with an exploration report

Exploration reports provide the fastest way to isolate LLM referral traffic. They display historical data without altering your GA4 property's default settings. This method is fully reversible, so you can safely build, test, and adjust your filters before committing to any permanent configurations.

Create the exploration report

1. Navigate to the Explore section in the left navigation menu of GA4.
2. Click the plus icon to create a new Blank exploration.
3. Name your exploration something clear, such as "LLM Traffic."
4. Import Session source as a dimension in the Variables column.
5. Import Sessions, Engagement rate, and Conversions as metrics.
6. Drag the dimension into the Rows setting and the metrics into the Values setting.

Apply a regex filter for AI referrers

To isolate traffic originating from AI assistants, apply a regular expression (regex) filter to the dimension. This weeds out standard search engines and social networks, leaving only the AI platforms you want to analyze, as outlined in a guide on VoltEdu.

In the tab settings of your exploration, scroll to the filters section and add a filter for the dimension. Set the match type to "matches regex" and enter your target AI referrers. This configuration is effective for verifying ChatGPT sessions, as well as applying the matching setup for Perplexity referrals and other models like Claude or Gemini.

Read the results

Once the filter is applied, the table displays only the sessions that match your AI referrer list. You will see how many sessions each AI tool generated, the engagement rate of those visitors, and whether they completed conversions. This dynamic view lets you analyze the performance of AI referrals without risk of corrupting your main reporting views.

Turn it into a permanent custom channel group

Creating a custom channel group persists your LLM traffic filters across the standard acquisition reports your team already checks. This method uses the same underlying logic as the regex filter, applied at the channel-definition level.

Once defined, LLM traffic automatically appears as its own row alongside categories like Organic Search, Direct, and Referral. This permanent structure is ideal for establishing a broader AI referral tracking setup. It lets you track visits from ChatGPT, Gemini, and other conversational engines without manual filtering every time you open GA4.

1. Navigate to the Admin settings in the lower left corner of your GA4 property.
2. Select Data display and click on Channel groups.
3. Click Create new channel group.
4. Define a new channel named "LLM Traffic" using the session source regex from your exploration filter.
5. Reorder your rules so the LLM group is evaluated before standard Referral traffic.
6. Save the group.

What to check once LLM traffic is visible

Begin by comparing its engagement rate and conversion performance against your traditional referral channels. AI assistants generally route users to highly specific internal pages—like detailed guides, documentation, or product pages—rather than your homepage. Analyzing which specific landing pages receive these sessions reveals exactly what content AI models use to satisfy user queries.

If you are driving visitors through a paid AI-traffic platform like PerkFuel, you can reconcile your GA4 data with your campaign settings. When setting up a project, you choose the target AI sources, landing pages, target countries, and delivery pace. With analytics tracking enabled, the visits in your PerkFuel dashboard should align closely with the corresponding referral sessions shown in Google Analytics.

Common mistakes when tracking LLM traffic in GA4

GA4 uses default channel grouping rules that can misclassify AI traffic. Without specific regex rules, traffic from Gemini often gets lumped into your standard organic search bucket because of its source domain structure. This is why Gemini traffic gets miscounted as Google organic, masking its actual performance.

Another gap occurs on mobile devices. When users click links within native AI assistant mobile apps, those apps sometimes open pages in restricted in-app browsers that strip referral data. These visits frequently register in GA4 as Direct / (none), which artificially lowers your measured AI traffic.

Do not treat your filters as a set-and-forget task. The AI landscape shifts rapidly as existing platforms change their referral domains and new assistant tools launch. To maintain accurate reporting, review and update your regex filters and custom channel definitions periodically.

To get a clear view of your AI-driven audience, set up your custom channel group today and reorder it to evaluate before standard referrals. Mobile in-app browsers will still occasionally strip referrer data, dropping some AI assistant sessions into your Direct traffic bucket. Set a recurring calendar reminder to update your regex filter list as new models launch and referral domains evolve.