How to Track Perplexity & ChatGPT Referral Traffic in GA4 (2026)

Somebody asked ChatGPT for "the best barefoot shoes for wide feet" this morning, got your brand recommended, clicked through, and bought. In your analytics, that sale is filed under Direct. No source, no channel, no credit.

That's the state of AI referral traffic in 2026: it's real, it's growing every quarter, and most ecommerce teams can't see it. This guide covers what GA4 now does automatically, where it still fails, and the setup that gets you as close to the full picture as the data allows.

What Counts as AI Referral Traffic?

AI referral traffic is any session that starts inside an AI assistant — ChatGPT, Perplexity, Gemini, Claude, Copilot — and lands on your site through a cited link or recommendation.

It behaves differently from search traffic. The assistant has already done the comparison shopping: by the time someone clicks through, they've read a synthesised answer about your product, your pricing, and your alternatives. These are mid-to-bottom-funnel visitors arriving on whatever page the assistant happened to cite — often a blog post or a deep PDP, not your homepage.

ChatGPT dominates the category with roughly 78% of global AI referral share, with Gemini overtaking Perplexity for second place in early 2026. But the split varies wildly by niche — check your own data before assuming.

Step 1: Check GA4's Native "AI Assistant" Channel

Since May 2026, GA4 automatically classifies traffic from recognised AI assistants — including ChatGPT, Gemini, and Claude — under a native AI Assistant channel. If you haven't looked since then, start here:

1. Open Reports → Acquisition → Traffic acquisition
2. Set the primary dimension to Session default channel group
3. Look for the AI Assistant row

If it's there and populated, you already have a baseline. But don't stop reading — the native channel only catches sessions that arrive with an intact referrer header, and that's where things get messy.

Step 2: Understand the Referrer Gap

Here's the part most guides skip: a large share of AI referral sessions arrive with no referrer at all. Statcounter data from March 2026 puts the figure between 35% and 70%, depending on the platform. Mobile apps, embedded browsers, and the assistants' own link-handling strip the referrer header — and those sessions land in Direct.

Practically, that means whatever number GA4 shows you for AI traffic is a floor, not the total. If your "Direct" channel has grown faster than your brand awareness plausibly has, AI assistants are a likely culprit.

A useful heuristic: segment Direct traffic by landing page. Direct sessions landing on deep informational URLs — a specific blog post, a comparison page, a PDP three levels down — almost never come from people typing the URL. They come from somewhere that stripped the referrer.

Step 3: Build a Custom Channel Group

The native channel is a good default, but a custom channel group gives you control over which platforms count and how they're ordered. In Admin → Data settings → Channel groups, create a channel with a regex match on session source:

chatgpt\.com|chat\.openai\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|copilot\.microsoft\.com|bing\.com/chat|deepseek\.com|grok\.com|meta\.ai|you\.com

That pattern covers the platforms responsible for the overwhelming majority of measurable AI referral traffic as of mid-2026.

One ordering rule that matters: place your AI channel above Referral in the channel group. GA4 assigns traffic to the first matching channel, so if Referral sits higher, your AI visits get lumped into generic referrals and you're back where you started.

Step 4: Watch What AI Visitors Actually Do

Once the channel exists, treat it like any other acquisition channel and ask the conversion questions:

• Which pages do assistants cite? (Reports → Engagement → Landing pages, filtered to your AI channel.) These are your de facto AI storefronts — whether you designed them as such or not.
• How does AI traffic convert vs organic? Early industry data says these visitors are higher-intent but less forgiving: they arrive pre-sold and bounce hard if the landing experience doesn't match what the assistant promised.
• Which products get recommended? The assistant controls the shortlist. If you're only cited for one product, that's your AI-channel catalogue today.

The Real Question: What Happens After the Click?

Tracking AI traffic is step one. The uncomfortable follow-up is that most of this traffic lands on pages built for human browsers coming from Google — not for a shopper who just read a synthesised recommendation and wants to act on it.

That's a post-click problem, and it's the same one paid social solved with campaign-specific micro-stores: match the storefront to the context that sent the visitor. An AI referral deserves a landing experience that confirms the recommendation, shows the cited product first, and gets out of the way.

That's what we're building with AI traffic storefronts — dedicated, catalogue-synced stores for the AI-agent channel, the same way Comet already does it for creators and campaigns.

Your AI channel is small today. So was paid social once. The brands that win it will be the ones who could see it first — set up the tracking this week, then book a demo when you're ready to give that traffic somewhere better to land.