How to track AI search traffic

Assistants have started sending real people to websites, and most analytics setups file that traffic under nothing in particular. So the first question leaders ask, how much of our traffic comes from AI, usually gets answered with a shrug. It does not have to be.

You track AI search traffic by isolating the referrals assistants send you. In GA4, build a segment or channel that matches the assistant domains, then watch which pages they land on.

There is one catch to be honest about from the start. This captures clicks from assistants, not citations inside AI Overviews, which send no referral. So referral tracking is half the picture, and the more important half is measured a different way.

What counts as AI search traffic

AI search traffic, in the sense you can actually measure, is the referral clicks that assistant products send to your site.

When someone reads an answer in ChatGPT, Perplexity, or Gemini and clicks a cited link through to you, that visit arrives as a referral from the assistant's hostname. That is the traceable part. It is real, it is growing, and almost nobody has it segmented.

What does not count here is the much larger volume of times an assistant describes or recommends you without anyone clicking. That is visibility, not traffic, and I come back to it below because it changes how you read the numbers.

Find the assistant referrals in GA4

Start by confirming the traffic exists. In GA4, look at your traffic acquisition report and add session source as the dimension, then look for the assistant hostnames.

The recognisable ones are chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, and claude.ai, with openai.com and bing.com appearing in some setups. You will usually find the numbers are small. That is expected. The point of the exercise is to see it cleanly and watch it grow, not to discover it is already big.

Keep the list under review. Assistants rename hostnames and new ones launch, so a check every quarter stops your tracking quietly going stale.

Build a segment or channel for AI traffic

Seeing it once is not tracking it. To follow AI search traffic over time, give it its own bucket.

The durable way is a custom channel group in GA4 with a channel defined by a regular expression matching the assistant roots, so any of those sources rolls up into one "AI Assistants" line you can trend and compare against Organic and Direct. A saved segment does the same job for ad-hoc analysis. A regex on the recognisable roots is easier to maintain than an exact domain list, because it survives small hostname changes.

Once the channel exists, AI traffic stops being something you go hunting for and becomes a line on the same reports you already read.

Watch which pages AI sends people to

The most useful thing this unlocks is not the total. It is the page breakdown.

Look at which landing pages your AI channel actually arrives on. That tells you which of your pages assistants are citing and sending people to, which is a different and sharper signal than which pages rank. A page pulling assistant referrals is one an assistant found worth citing, which is direct feedback on what your citable content looks like.

Cross-referenced with your rankings, it also helps you read the gaps: pages that rank well but pull no AI traffic are candidates for the extractability work that earns citation.

What AI search traffic tracking misses

Here is the limitation to hold onto, because it is the difference between a useful metric and a misleading one.

Referral tracking only sees clicks. Google's AI Overviews, the largest AI surface most of your buyers touch, sends no referral when it cites you inside the answer, and any click it does pass arrives as ordinary Google organic, not a distinct AI source. So a low AI referral number does not mean you are absent from AI answers. It means referral is not measuring the part that matters most.

That is why tracking traffic is only ever half the job. The other half is measuring visibility directly: running a fixed set of buyer prompts across the assistants and recording whether you are cited, mentioned, or absent. The full method is in measuring brand visibility in AI answers, and for the Google surface specifically, how to track AI Overviews covers presence and citation, which referral traffic never will.

If your referral traffic is flat while your rankings hold and clicks fall, that is not a tracking gap. It is the rankings-hold-traffic-falls pattern, and it is a visibility problem, not a measurement one.

The takeaway

Track AI search traffic by isolating assistant referrals in GA4: identify the hostnames, roll them into a dedicated channel or segment, and watch which pages they land on. Then hold the number in perspective. It captures assistant clicks, not AI Overview citations, so pair it with a fixed prompt set that measures the visibility referral traffic cannot see.

If you want both halves measured together, the clicks and the citations, that is what an AI visibility audit provides.

This article is part of the SEO in the AI Era: The Complete Guide guide.