How to measure AI traffic from ChatGPT and Perplexity in Google Analytics 4

If you want to track traffic from ChatGPT or Perplexity in Google Analytics 4, you can create a custom segment or exploration that filters for the source domains chatgpt.com, perplexity.ai, gemini.google.com, and copilot. This will show you the clicks that AI search engines actually send to your website. What you won't see, however, is how often you are cited without anyone clicking. That is the most important limitation to understand before interpreting the numbers.

AI traffic refers to visitors who click through from an AI service to your site. AI visibility is how often AI engines mention and cite you in their answers, regardless of clicks. GA4 measures the former, not the latter.

This guide is a follow-up to the more in-depth review of what the GA4 report actually shows and doesn't show. Here you will find the exact click paths, a copyable regex, and the setup you can reuse.

How do I see traffic from ChatGPT?

ChatGPT traffic appears in GA4 as referral traffic from chatgpt.com. Here is how to find it quickly:

1. Open GA4 and go to Reports in the left menu.

2. Select Acquisition, then Traffic acquisition.

3. Click the dropdown above the table (default is "Session default channel group") and switch to the dimension Session source / medium.

4. Use the search box above the table and search for chatgpt.

You will typically see rows like chatgpt.com / referral. These are the visitors who clicked on a link in a ChatGPT answer.

If you want to include Perplexity, Gemini, and Copilot in the same view, it is better to build a combined segment instead of searching for one domain at a time. That comes next.

An important clarification: ChatGPT's search function is largely based on Bing's index, while OpenAI is also developing its own crawler (OAI-SearchBot). Appearing in ChatGPT is therefore partly related to how you are indexed in Bing, but not exclusively.

Setting up a segment for AI traffic in GA4

You get the real picture of AI traffic by combining all the major search engines in one filter. In GA4, the best way to do this is in an Exploration (Explore), because that is where you can use segments freely and build tables across sources.

Here is how to set it up:

1. Go to Explore in the left menu and create a Blank exploration.

2. Under Segments, click + to create a new segment. Choose Session segment.

3. Set the condition to the dimension Session source and select the operator matches regex.

4. Paste the regex below.

5. Drag in dimensions like Session source / medium and Landing page, and metrics like Sessions, Engaged sessions, and Conversions, to see both the volume and quality of your AI traffic.

Copyable regex for AI sources

chatgpt\.com|perplexity\.ai|gemini\.google\.com|copilot

chatgpt\.com|perplexity\.ai|gemini\.google\.com|copilot

chatgpt\.com|perplexity\.ai|gemini\.google\.com|copilot

This captures the four most common response engines in the Norwegian B2B context today. Some tips:

• Periods are escaped with \ so they match a literal period, not any character.

• copilot has no domain because Microsoft Copilot traffic can come through several hosts. Check your own source rows and tighten up if you see noise.

• If you want to add more, use | (or symbol) between each: for example, add |you\.com at the end.

Save the segment, and you can reuse it in all explorations. You can also create a corresponding channel group or a comparison filter in standard reports with the same logic, but exploration offers the most flexibility.

Why GA4 does not show the full AI picture

Here is the distinction that determines how you should interpret the numbers: GA4 only counts clicks with a recognized referral. This means the report has three major blind spots.

Citation without click. When ChatGPT or Perplexity summarizes the answer with you as the source, but the user gets what they need without clicking, nothing happens in GA4. You were visible, but invisible in the analysis.

In-app and missing referrer. Traffic from mobile apps or services that do not send a referral value often ends up under Direct instead of as an AI source.

Google AI Overviews and AI Mode. Clicks from Google's AI answers are usually counted under Organic Search, not as a separate AI channel. GA4's built-in AI channel does not capture these either.

The consequence is simple: use GA4 to measure AI traffic, but do not confuse it with AI visibility. Visibility, meaning how often you are actually cited, must be measured using other tools and methods. That is the entire point of the work described in the flagship guide to AI visibility for Norwegian B2B.

Why is this distinction so important? Because the search engines overlap surprisingly little. An analysis by Profound of 100,000 prompts on ChatGPT and Perplexity found only an 11 percent domain overlap, and between 6 and 16 percent among the major search engines. Being cited in one engine does not mean you are visible in another. The click figures in GA4 tell you nothing about this coverage.

How to build on the numbers

Once the segment is set up, you can use it actively:

• Track the trend over time. Compare the AI segment month over month. You are looking for direction, not absolute numbers, since volumes are often small at the start.

• Look at the landing pages. Which pages are getting AI clicks? This tells you which content the search engines already deem worthy of answering with.

• Measure quality, not just volume. Add engaged sessions and conversions to the table. AI traffic is often low volume, but can be highly intentional.

To actually increase both traffic and visibility, you must work on the content itself. Research on generative engine optimization (GEO) from Princeton and the Allen Institute shows that content with citations, statistics, and clear source references becomes significantly more visible in AI answers—up to a 40 percent increase in visibility—while keyword stuffing has little effect. Specifically, the study found that citations increased visibility by 41 percent, statistics by 31 percent, and source references by 30 percent. This is the kind of work we do in our AI visibility service.

Frequently asked questions

How do I see traffic from ChatGPT in Google Analytics 4?

Go to Reports, Acquisition, Traffic acquisition, change the dimension to Session source / medium, and search for chatgpt. You will then see rows like chatgpt.com / referral, which are clicks from ChatGPT answers. For a complete picture of all AI traffic, you should instead create a segment using regex in an exploration.

Which regex do I use to capture all AI traffic in GA4?

Use chatgpt\.com|perplexity\.ai|gemini\.google\.com|copilot on the Session source dimension with the operator "matches regex". It covers ChatGPT, Perplexity, Gemini, and Copilot. Extend with | in front of more domains if you want to add sources like you\.com.

Why do I see so little AI traffic in GA4 even though I know I am being mentioned?

Because GA4 only counts clicks with a recognized referral. If you are cited without anyone clicking, or if the traffic comes from an app with no referrer, it will not appear as an AI source. This is the difference between AI traffic (clicks) and AI visibility (citations). GA4 only measures the former.

Are Google AI Overviews counted as AI traffic in GA4?

No, usually not. Clicks from Google's AI Overviews and AI Mode typically end up under Organic Search, not as a separate AI channel. This means GA4 underestimates how much of your search traffic actually passes through an AI-generated response.

How do I measure AI visibility when GA4 doesn't?

Visibility, i.e., how often and how you are cited in AI answers, must be measured using dedicated tools that check responses to fixed prompts over time, across multiple engines. Since the engines overlap very little (Profound found only an 11 percent domain overlap between ChatGPT and Perplexity), you must measure each engine individually. This is the core of the flagship guide to AI visibility.

Should I create a custom channel group for AI in GA4?

You can, and it makes for cleaner standard reports. However, a custom channel group only affects data going forward and cannot be used retroactively in the same way as a segment in an exploration. Start with the segment, and consider a channel group once you know that your AI traffic is large enough to monitor regularly.