How to Track AI Referral Traffic in GA4

Why is AI referral traffic so hard to see in GA4 by default?

Here's the frustrating part: when someone asks ChatGPT or Perplexity a question and clicks through to your site, that visit rarely announces itself. Out of the box, GA4 files those sessions under the generic Referral channel — lumped in with forums, newsletters and every other site that links to you. So you know traffic is arriving, but you can't tell that an answer engine sent it. For founders trying to prove that their GEO work pays off, that's a real blind spot.

It gets thornier. A large share of AI visits arrive with no referrer at all — often well over a third, and sometimes as much as 70% — because they come from desktop and mobile apps that strip the referring header. Those land in Direct traffic, indistinguishable from someone typing your URL. No configuration can rescue every one of them. The goal isn't perfect capture; it's isolating the measurable majority so you can watch the trend and make decisions with real numbers instead of hunches.

What changed with GA4's native AI Assistant channel in 2026?

In May 2026, Google added a native AI Assistant channel to GA4's Default Channel Group. When an incoming session's referrer matches a recognized AI domain, GA4 now stamps it with the medium ai-assistant and files it under AI Assistant automatically — no setup required. For anyone who dreaded building regex, that's a welcome shortcut, and it's the easiest way to start reading ongoing AI traffic at a glance.

There's a catch, though. The native channel only recognizes a short list of platforms — Google's documentation has named the likes of ChatGPT, Gemini, Claude, Copilot, Deepseek and Grok, and the exact roster keeps shifting. Notably, Perplexity has been absent, so those sessions still fall into Referral. Just as importantly, the native channel doesn't backfill: it only tags traffic going forward. If you want your historical AI visits attributed too, you still need a custom channel group. The smart move in 2026 is to run both in parallel.

Which AI domains should your GA4 rules match?

Accurate tracking starts with a clean, current list of source domains. Each answer engine sends traffic from one or more hostnames, and the big five cover the overwhelming majority of measurable AI referrals today. The table below groups the platforms most worth capturing as of mid-2026 — together they account for well over 99% of trackable AI referral traffic.

One caution on Google's own products: AI Overviews and AI Mode in Google Search count as Organic Search, not AI referral, so you won't find them here no matter how you configure things. And resist the temptation to write a lazy catch-all like anything containing "ai" — it will sweep in email providers, retailers and unrelated domains, quietly inflating your numbers and destroying trust in the report.

How do you build a custom AI channel group step by step?

Open GA4, click the Admin gear at the bottom left, then head to Channel Groups under Data Display. Choose Create New Channel Group (or edit an existing custom one). Add a new channel — call it "AI Assistants" — and define a rule based on Session source that matches your AI domains. A clean, anchored regex beats a stack of overlapping rules: try ^(chatgpt\.com|chat\.openai\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|copilot\.microsoft\.com|deepseek\.com|grok\.com|meta\.ai|you\.com)$ and extend it as new engines appear.

Now the step that trips almost everyone up: rule order. GA4 evaluates channel rules top-down and stops at the first match. If your AI rule sits below Referral, AI sources hit Referral first and get miscategorized — and you'll swear your regex is broken when it isn't. Drag the AI Assistants channel above Referral so it's evaluated first. The beauty of a custom group is that it applies retroactively: the moment you save it, GA4 rebuckets your historical source and medium data, revealing AI traffic as far back as your records go.

How do you report on and sanity-check the numbers?

Once the channel group exists, you can slice it anywhere GA4 lets you pick a dimension. For a quick pulse, open Reports and switch the channel dimension to your custom group. For deeper analysis, go to Explore, create a Free-form report, add Session source as a dimension and Sessions plus Engaged sessions as metrics, then filter Session source with your AI regex. That view lets you compare engines head to head — which one actually drives engaged visits, not just clicks.

Always pressure-test what you see. If a freshly built group still shows AI traffic under Referral, the culprit is nearly always rule order, not your pattern. Watch engagement and conversions per engine, not just raw sessions — a small stream of highly qualified visitors from an answer engine can outperform a larger, shallower channel. And remember the Direct-traffic caveat: your reported AI numbers are a floor, not a ceiling, because referrer-stripped app visits stay hidden.

The payoff is a defensible answer to the question every founder eventually asks: are AI answer engines sending real people, or is this hype? With this setup, you can point at a chart instead of guessing.

Native channel or custom group — which should you use?

You don't have to choose. The native AI Assistant channel is the low-effort way to keep an eye on ongoing AI traffic, and it requires zero configuration. The custom channel group is what gives you historical backfill, full control over which domains count, and coverage for stragglers like Perplexity that Google may not include. Used together, they cross-check each other and close the gaps.

If you only have five minutes, lean on the native channel today and schedule time to build the custom group this week. Whichever path you take, treat the domain list as a living document — Grok and DeepSeek were rounding errors in 2025 and are measurable sources now. Review your regex quarterly so a new engine doesn't quietly slip into Referral and rob your AI channel of credit.

How does AI-traffic tracking fit a wider organic-marketing strategy?

Measurement only matters if it feeds action. Once you can see which answer engines cite and refer you, you know where your content is winning — and where it's invisible. That's the loop GEO is built on: publish genuinely useful, well-structured content, watch which engines pick it up, then double down on the formats and topics that earn citations across ChatGPT, Claude and Perplexity in every language your buyers use.

This is exactly where an organic-marketing autopilot earns its keep. Artiql produces multilingual SEO and GEO articles designed to rank on Google and get quoted by AI answer engines, pairs each one with an AI video that flows from YouTube to Instagram and TikTok, and routes everything through a review queue into a headless CMS on your own domain. You keep the strategy; the busywork runs itself. If you'd like to see how it maps to your funnel, book a demo and we'll walk through it with your own topics.

Frequently asked questions