Insights / Threads
How to measure traffic to your site from AI in GA4
To measure traffic from ChatGPT, Gemini, or Perplexity in GA4, start by reviewing referral sources and creating a specific channel group for AI. Measurement will not be perfect, but it lets you separate sessions, landing pages, and conversions coming from generative environments.
How to measure traffic from ChatGPT, Gemini, or Perplexity in GA4
Measuring traffic from ChatGPT, Gemini, or Perplexity in GA4 means identifying visits that arrive from domains associated with AI tools and separating them from the rest of your channels. It is not analytics magic, but it does require organizing sources properly.
GA4 can only measure what arrives with detectable signals. If a visit comes from chatgpt.com, perplexity.ai, or gemini.google.com, you can see it as a referral or session source. If it arrives without a referrer, it may end up mixed into direct traffic.
That is why measurement should be read as a useful approximation, not as a complete picture of your entire AI visibility. It helps you see part of the impact, but it does not replace a qualitative review of mentions, citations, and positioning in generated answers.
Where to find AI traffic in GA4
The first place to review AI traffic in GA4 is the acquisition reports. You can go to traffic acquisition and analyze dimensions such as session source/medium, session source, or default channel group.
There, it is worth looking for domains associated with specific tools. For example: ‘chatgpt.com’, ‘chat.openai.com’, ‘perplexity.ai’, ‘gemini.google.com’, ‘copilot.microsoft.com’, or ‘claude.ai’, depending on the environments you want to monitor.
The basic reading is straightforward: if those sources appear, you already have signals of referral traffic from AI. The interesting part comes next, when you look at which pages receive those visits, how users behave, and whether they end up generating relevant events.
How to create an AI channel group in GA4
Creating a custom channel group for AI helps avoid mixing this traffic with traditional referral. In GA4, you can configure your own grouping using rules based on source, medium, or referral domain.
A practical rule is to include sources that match AI tool domains. For example, you can group patterns such as ‘chatgpt’, ‘openai’, ‘perplexity’, ‘gemini’, ‘bard’, ‘copilot’, ‘claude’, or ‘poe’, adjusting the list according to your market and real data.
You do not need to find a perfect regex on day one. It is better to start with a reasonable list, review it periodically, and add new sources as they appear. The ecosystem changes quickly, and some names or domains evolve.
What metrics to review for AI traffic
Sessions are only the entry point. To understand whether traffic from ChatGPT, Gemini, or Perplexity creates value, it is worth looking at landing pages, engagement time, key events, conversions, forms, scroll, and subsequent paths.
It is also useful to separate informational content from business pages. If AI tools send traffic to an article, that may indicate editorial citability. If those users also move toward services, contact, or case studies, the signal starts to have more commercial value.
Our recommendation is to create a specific GA4 exploration with session source, landing page, key events, and conversions. This lets you see not only which tool brings traffic, but what type of content is working as an entry point.
What limitations GA4 has with ChatGPT, Gemini, and Perplexity
GA4 cannot attribute the entire impact of AI. Some traffic may arrive as direct if the referrer is not preserved. Some interactions happen inside the tool itself, without a visit to your website. And presence in generated answers can influence decisions even when it does not produce an immediate click.
It is also important not to put everything into the same bucket. Perplexity often works differently from ChatGPT, Gemini can mix with the Google ecosystem, and AI Overviews are not always separated from organic search. If you simplify too much, the reading becomes comfortable but imprecise.
The sensible approach is to use GA4 as one measurement layer, not as the only source of truth. For GEO, you need to combine web analytics with manual review of answers, citations, competitors, and mention quality.
How to interpret AI traffic in GA4
Interpreting AI traffic in GA4 requires looking at trend, quality, and context. A one-off spike may come from an isolated citation. Sustained growth across several pages may indicate that your content is becoming useful as a source.
The most valuable step is to detect patterns: which topics attract visits from AI, which URLs appear most often, which tools generate better engagement, and which content converts better after the initial visit.
If you do that reading monthly, you start building a much more useful picture than simply saying “we got traffic from ChatGPT.” You can know which pieces are being used, what intent users bring, and where it is worth strengthening content or internal links.
Frequently Asked Questions
Yes, but with limitations. GA4 can show sessions that arrive from referral domains such as chatgpt.com, perplexity.ai, or gemini.google.com. The problem is that not all AI traffic arrives with a clean referrer: some of it may appear as direct, generic referral, or organic traffic.
You can start in the acquisition reports, reviewing source/medium or session source. Look for domains such as chatgpt.com, chat.openai.com, perplexity.ai, or gemini.google.com. After that, it is worth creating an exploration or a specific channel group to isolate this traffic.
Yes. Creating a custom channel group helps avoid mixing visits from AI assistants with traditional referral or direct traffic. This lets you analyze sessions, conversions, landing pages, and changes over time with more clarity.
Not separately or perfectly. In many cases, traffic related to Google results may appear within organic search. That is why it is important not to confuse traffic from Gemini or conversational tools with visibility inside Google generative results.
Beyond sessions, it is worth looking at landing pages, key events, conversions, engagement, associated queries when available, and changes by source. The question is not only how much traffic arrives, but whether that traffic lands on strategic content and generates useful signals.
To dig deeper into this topic
