Measuring AI Search: How to Track What Came From ChatGPT

There is a reasonable chance that AI search is already sending you customers. There is also a reasonable chance you have no idea, because you are not measuring it.

This is not a niche problem. AI referral traffic converts at roughly 9x the rate of Google organic traffic. If that traffic is hitting your site and being attributed to "direct" or "organic," you are making resource decisions on incomplete data.

Here is how to close the gap.

What GA4 Already Captures

The good news: GA4 already tracks some AI referral traffic. When a user clicks a link in a ChatGPT response and lands on your site, the referral source is recorded as chatgpt.com. Same for perplexity.ai and gemini.google.com (when the user clicks through from Gemini's cited sources).

The bad news: most analytics dashboards are not set up to surface this. It gets buried in the "other" referral bucket or treated as general direct traffic.

To find it today, open GA4 and navigate to Reports > Acquisition > Traffic Acquisition. Filter by Session source and search for:

• chatgpt.com
• perplexity.ai
• gemini.google.com
• claude.ai
• copilot.microsoft.com

If these show up at all in your data, you are already getting AI referral traffic. The number is almost certainly understated, but it is a real baseline.

Building a Proper AI Referral Segment

Rather than checking manually each time, build a persistent segment in GA4 that captures all AI referral sources in one view.

In GA4, go to Explore > Create a new exploration. Add a segment condition: Session source contains any of the following domains. Add chatgpt.com, perplexity.ai, gemini.google.com, claude.ai, copilot.microsoft.com, and bing.com (Bing AI). Save this as a reusable segment called "AI Search Referrals."

Apply this segment to your conversion reports. You will see sessions, conversions, and revenue attributed to AI platforms side by side. Compare the conversion rate against your Google organic segment. The difference will clarify why this measurement matters.

UTM Parameters for Controlled Campaigns

If you are actively trying to drive traffic from AI platforms (through content placed on high-authority sites that AI tends to cite, or through directories AI references), UTM parameters let you track what is working.

Structure your UTMs like this:

• utm_source: the AI platform (chatgpt, gemini, perplexity)
• utm_medium: ai-referral
• utm_campaign: the specific initiative (geo-optimization, review-campaign, etc.)

This is most useful when you are publishing content to external sources and want to attribute downstream AI-driven traffic back to those content investments.

The Attribution Gap You Cannot Fully Solve Yet

Here is the harder problem: the majority of AI-influenced traffic is not directly attributable.

The pattern works like this. A user asks ChatGPT for a recommendation. ChatGPT names your business. The user does not click a link in ChatGPT. Instead, they open a new browser tab and Google your name. They land on your site from Google organic, or type your URL directly. GA4 credits Google or direct. ChatGPT gets no credit.

This is called dark AI traffic. It is invisible in your current analytics. The AI recommendation was the trigger. Your analytics do not see it.

The scale of this is unknown, but directionally significant. Think about your own behavior. When ChatGPT recommends a restaurant, you probably do not click through from the ChatGPT interface. You Google the name or open their site directly. That branded search or direct visit is AI-influenced but not AI-attributed.

There is no clean solution to this yet. Some proxies help:

• Branded search volume. Track your brand's Google Search Console impressions over time. An increase in branded queries after you improve AI visibility is a signal, not proof, that AI recommendations are generating awareness.
• Direct traffic trends. A rising direct traffic baseline alongside increased AI mentions suggests dark AI traffic is accumulating. Correlation, not causation, but worth watching.
• Customer surveys. At booking or purchase, ask "How did you first hear about us?" Add "AI assistant (ChatGPT, Gemini, etc.)" as an option. Self-reported but directionally useful.

What to Do This Week

Three steps, in order:

1. Check GA4 for existing AI referral traffic. You may already have data you are not looking at.
2. Build the AI referral segment. Make it a permanent fixture in your reporting, not a one-time check.
3. Add "AI assistant" to your customer acquisition survey if you have one. Start capturing self-reported data now. It will be more useful in six months than it is today.

The measurement infrastructure for AI search is early. But the gap between what is measurable today and what businesses are actually measuring is large. Closing that gap is a competitive advantage while most of your competitors are still ignoring it.