AI assistants are already sending you customers. When ChatGPT cites your page in an answer and the user clicks the source, when Perplexity links you as a reference, when someone reads a Gemini summary and then visits your site, that is real, often high-intent traffic. The problem is that if you open GA4 and look for it, most of it is not where you expect, and a large share of it is invisible entirely. Before you can make the case that AI visibility is worth investing in, you have to be able to see the traffic it produces. Here is how, and why the number you get will still be lower than the truth.

Why GA4 Undercounts It by Default

Two things conspire to hide AI referral traffic.

First, a lot of it lands in GA4 as "Direct / (none)." When a user clicks a link inside a native app, the ChatGPT app, the Perplexity app, an AI feature baked into another tool, the referrer information is frequently stripped before it ever reaches your site. GA4 has nothing to attribute the visit to, so it files it under direct, the same bucket as someone typing your URL. Your AI traffic is sitting inside your direct traffic, unlabeled.

Second, when a referrer does come through, GA4 does not group AI sources into a tidy channel. A visit from chat.openai.com or perplexity.ai shows up as an individual referral source, scattered and easy to miss, rather than rolled into anything called "AI Assistants." Out of the box there is no such channel, so unless you build one, the traffic never aggregates into a number you would actually notice.

Add those together and the default GA4 view makes AI traffic look far smaller than it is.

That is not a bug you can fully fix, but you can get much closer to the real figure.

Step One: Isolate the Referrers You Can See

Start with the traffic that does carry a referrer. Build an exploration in GA4 filtered to session source containing the known AI hosts, chatgpt.com and chat.openai.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, and the handful of others relevant to your audience. That gives you a floor: the AI traffic GA4 can positively identify.

To make this durable instead of a one-off report, create a custom channel group that captures those same hosts and labels them as AI Assistants. Once that channel exists, AI traffic aggregates automatically in your standard reports going forward, and you stop having to rebuild the filter every time. This is the same discipline behind getting any attribution right, which we get into in GA4 attribution models and why most marketers misuse them.

Step Two: Recover What Is Hiding in Direct

The referrer-visible traffic is only the part GA4 can see. To estimate the rest, look at your Direct / (none) traffic for the fingerprints of AI referrals: landing pages deep in your site rather than the homepage, visited by new users, with engaged behavior. People do not usually type a deep article URL from memory. A rising tide of direct traffic to specific internal pages, especially pages that answer a clear question, is very often AI and answer-engine referrals arriving without a referrer.

The most reliable way to firm this up is UTM tagging on any link you yourself place into content that AI systems ingest, and watching Search Console and server logs for the crawlers those systems use. You will never get the direct-traffic estimate to be exact, and any tool that promises precise, comprehensive AI-referral numbers is overselling. Treat it as a directional signal, not a hard count, and be honest about that when you report it.

Step Three: Judge the Traffic on Behavior, Not Volume

Once you can see it, resist the urge to compare its raw volume to Google organic and conclude it is trivial. AI referral traffic is usually smaller in count but qualified in intent, because the user already got context from the AI and clicked through specifically to go deeper or to act. Look at engagement rate, pages per session, and conversions for your isolated AI segment. If those metrics are strong, a small traffic number represents a large opportunity, and it will grow.

This is also the measurement half of a strategy whose other half is earning the citations in the first place. Being the source AI systems quote is what generates this traffic, and the tactics for that live in the zero-click search playbook. Measurement without the visibility work shows you an empty channel; visibility work without measurement means you cannot prove it paid off.

What to Do With the Number

The goal of tracking AI traffic is not a vanity metric. It is to answer one question: is the effort you are putting into being cited by AI systems producing qualified visitors and conversions? Once you have an AI Assistants channel and a reasonable estimate of the direct-traffic portion, you can watch it as a trend, tie it to the content that earns citations, and make an evidence-based call about how much to invest.

For the mechanics of the reports and channel groups, Google keeps the current documentation in the Analytics Help center, and it changes often enough that it is worth checking against the live product rather than a year-old tutorial. The channel is young and the tooling is immature, but the traffic is real, and the businesses that start measuring it now will be the ones who can prove its value while everyone else is still calling it direct.