How to Track LLM Traffic in Google Analytics (Without Losing Your Mind)

Large language models (LLMs) like ChatGPT, Claude, and Perplexity are sending real traffic to your website. The problem? Your analytics platform is treating it like a mystery guest at a wedding—dumping it into the “referral” bucket and moving on.

If you’re seeing weird spikes in traffic or your ROI calculator is suddenly getting love from nowhere, congrats—you’ve been LLM’ed.

Here’s how to track that traffic properly using GA4 and Looker Studio (without rage-quitting your analytics setup).

Why Should You Care About LLM Traffic?

AI chatbots are now part of the modern discovery journey. Whether someone searches in ChatGPT or clicks through from a Claude response, those sessions are real—and they deserve better than getting tossed into “miscellaneous.”

If you’re a B2B marketer who actually wants to:

• Know what content the robots are recommending
  
• Attribute traffic that doesn’t show up in your CRM
  
• Optimize for the AI overlords (before your competitors do)
  

…then yeah, this matters.

Step 1: Create a Custom Channel Group in GA4

GA4 out of the box? Useless for this. You’ll need to DIY a custom channel group to catch traffic from known LLM sources.

How to do it:

1. Head to Admin > Data Display > Channel Groups
   
2. Add a custom channel group called “LLMs” (or go wild and name it something sassier)
   
3. Add a regex filter that captures source domains. We used:

^.*\.openai.*|.*copilot.*|.*gpt.*|.*neeva.*|.*writesonic.*|.*nimble.*|.*perplexity.*|.*outrider.*|.*bard.*|.*gemini.*|.*edgeservices.*$

4. Reorder it to the top of the list. GA4 processes these top-down. If “referral” is ahead of your new rule, you lose.
   

Remember: This only works going forward. It won’t backfill past data. Thanks, Google.

Step 2: Get Historical Data with Looker Studio

Want to see what the bots have already blessed you with? Enter Looker Studio (or any decent BI tool).

Here’s what we do:

• Create a filter using the same regex to find past LLM sessions
  
• Build a simple report showing source, engaged sessions, and landing pages
  
• See which content is getting picked up and why
  

Fun fact: Our top-performing LLM page was a no-frills ROI calculator blog. 

Step 3: Maintain Your Regex (Yes, Really)

Regex is powerful. It’s also temperamental. If ChatGPT changes their domain (and they will), your whole filter falls apart faster than your paid search attribution.

So:

• Review source domains every month
  
• Test any changes before shipping them
  
• Don’t try to be too clever. Clean, simple regex wins every time
  

If you end up capturing chat.unicorns.ai because of sloppy syntax, don’t say we didn’t warn you.

Job Done. For now.

AI tools are already reshaping how people find and engage with your content. If your analytics setup doesn’t track that traffic, you’re missing the plot—and the pipeline.

At Airfleet, we build and optimize B2B websites that perform under pressure (and under AI scrutiny). Want your site to actually convert when it shows up in ChatGPT? Let’s talk.