How to measure traffic and referrals from AI models

Traffic from AI models can partly be tracked in standard analytics tools by filtering referral sources by domains such as chat.openai.com, claude.ai and gemini.google.com, but most of the exposure in AI answers never produces a click to measure. That is why you need two different approaches: technical tracking of what is clickable, and direct measurement of the answers themselves for the rest.

What you can actually track in Google Analytics

When a user clicks a link inside a ChatGPT or Perplexity answer, they are often sent along with a referrer domain you can filter on in your analytics tool. Create a dedicated segment or channel group for known AI domains (chat.openai.com, chatgpt.com, claude.ai, gemini.google.com, perplexity.ai) so this traffic stands out from regular organic search in your reports. This gives you a minimum figure, never the full picture.

Why most of it remains invisible

Large parts of the exposure in an AI answer never lead to a click at all. The user gets their answer directly in the conversation, sees that your business is mentioned or recommended, and never visits the website to confirm it. This is often called zero-click exposure, and it is the norm rather than the exception in generative answers. That means even a solid AI visibility strategy can produce minimal measurable traffic growth, while still actually driving new customers who never clicked through.

Concrete steps for better tracking

• Use UTM parameters on links in content specifically written to be cited (guides, product pages, pricing pages), so clicks that actually come via AI tools are easier to identify even when the referrer is missing.
• Ask new customers directly how they found you. A simple field in the contact form or a question during the sales conversation captures referrals your analytics tool never sees.
• Keep an eye on brand searches in Google Search Console. An increase in searches for your business name can indicate that someone first heard about you through an AI model and then googled to confirm.
• Set up alerts for unusual traffic spikes without a clear source, these are often worth investigating manually for AI referral.

Measure the answers, not just the traffic

The most precise way to measure AI visibility is not to wait for traffic that may never come, but to measure the answer itself directly: are you mentioned, how often, and with what description. This is the same principle as in a practical AI visibility audit, just done repeatedly and systematically rather than as a one-off check.

Combine both methods over time

The best insight comes from combining technical traffic tracking with direct answer measurement, and viewing them together over several months. If exposure in the answers increases at the same time as brand search rises, that is a strong indication that AI visibility is actually converting, even without clickable proof.