How to Measure AI Referral Traffic in GA4 and Shopify

GA4 gave AI traffic its own row on May 13, 2026, when Google added an AI Assistant channel that classifies visits from ChatGPT, Gemini, Copilot and similar assistants with no setup. That sounds like the measurement problem is solved. It is not. The channel counts the floor of your AI traffic, not the ceiling, and it leaves your biggest AI source hiding inside Organic Search.

Here is the setup a lean team can run to see the real number, split it from Google Shopping, and reconcile it with Shopify.

What GA4 changed on May 13, 2026

GA4 added an AI Assistant channel to the Default Channel Group. When Google detects a referrer matching a recognized AI assistant, it assigns ai-assistant as the medium and groups the session under AI Assistant, with no regex and no custom setup, as Search Engine Journal reported at the time.

Google's default channel definitions describe AI Assistant as traffic from sources like ChatGPT, Gemini, DeepSeek, Copilot or Grok, and state that it excludes Google's AI Overviews and AI Mode. Two limits matter for retailers. The change is not retroactive, so May 13 is your baseline and everything before it stays miscategorized. And the channel only catches visits that carry a referrer header.

Why the native channel undercounts your real AI traffic

A meaningful share of AI-originated visits arrive with no referrer at all and land in Direct. Mobile apps, in-app browsers and agent browsers commonly strip the referrer, so a shopper who taps a product link inside the ChatGPT app can show up as if they typed your URL from memory. Google has not published a figure for this. SEO practitioners estimate the share at 20 to 40 percent of AI visits, and your own tagged-link data is the only way to know your number.

The second blind spot hits harder for commerce. Google's same channel documentation places AI Overviews and AI Mode inside Organic Search. For a retailer, Google's AI surfaces are often the largest AI discovery channel of all, and GA4 files them in a bucket you already think you understand.

The practical consequence: report AI traffic straight off the native channel and you will understate it, misattribute a chunk to Direct, and fold Google's AI discovery into Organic. Three different errors, all making AI look smaller than it is.

How to build a custom AI channel group in GA4

A custom channel group closes the gaps the native channel leaves and lets you fold in sources Google has not added yet. Google's custom channel groups documentation allows two custom groups per standard property, each with its own ordered rules. Here is the build.

Go to Admin, then Data Settings, then Channel Groups, and click Create new channel group. You cannot edit the default group, so this is a parallel view.
Add a channel named AI Search. Under Channel Conditions, choose the Source dimension and select matches regex.
Use a starting pattern that captures the main assistants: ^(chatgpt|perplexity|gemini|copilot|claude|openai\.com|poe\.com) and extend it as new surfaces appear.
Drag the AI Search channel above Referral and Organic Search. GA4 evaluates channels top-down, so if AI sits below Referral, those sessions get claimed by Referral first.
Save the channel and the group. View it under Reports, then Acquisition, then Traffic Acquisition, switching the dimension dropdown to your custom group.

This still will not recover referrer-less Direct sessions or Google AI Mode traffic, but it stops leaking the referrers you can see into the wrong buckets.

How to close the app and agent browser gap with UTMs

The referrer-less Direct problem is not fully solvable in GA4 alone, but you can tag the traffic you control. If you syndicate a product feed to AI shopping surfaces, append UTM parameters to the destination URLs so the click lands with explicit attribution instead of a stripped referrer. Where each signal lands by default, and what to do about it:

ChatGPT, Gemini, Copilot or Claude web referral: lands in AI Assistant. Confirm it with the custom channel group.
ChatGPT app and agent browsers with no referrer: lands in Direct. Tag feed and shared URLs with UTMs.
Google AI Mode and AI Overviews: lands in Organic Search. Segment by landing page and watch the query mix.
Perplexity and smaller assistants: lands in Referral or Direct. Add them to the custom regex and tag what you can.

A consistent scheme such as utm_source=chatgpt and utm_medium=ai_feed turns an otherwise invisible app click into a labeled session. Feeds are where AI discovery most often becomes a visit, so the feed that makes you visible in AI answers is the same feed that makes you measurable.

How to split AI referrals from Google Shopping

The question we hear most after the channel is built is how to tell AI traffic apart from Google Shopping and Merchant Center traffic. Google's own Merchant Center tracking guidance gives the rules. Shopping ads tracked through Google Ads auto-tagging arrive as source google, medium cpc, so they sit in Paid Shopping or Paid Search, not with AI. Free listings arrive as source google, medium organic, which puts them in Organic Search, the same bucket as Google AI Mode.

Two moves separate them. Turn on Merchant Center auto-tagging for unpaid traffic, which labels free listing clicks as Shopping free listings in GA4's source platform dimension, or use distinct link attributes in the feed with your own UTM values.

Then build an exploration on Organic Search with source platform and landing page as dimensions. Free listing clicks carry the label, AI Mode and AI Overview clicks do not, and both tend to land on product pages rather than category or blog pages. The product feed side of this, from auto-tagging to link attributes, is covered on our product feeds service page.

How to reconcile GA4 with Shopify analytics

Shopify's own analytics and GA4 will not match, and that is expected. Shopify attributes sessions and orders using its first-party data and its own last-click model, while GA4 uses its channel waterfall and its own attribution windows. Neither is wrong. They answer different questions.

Use Shopify to confirm orders and revenue by source, since it sees the checkout GA4 does not. Use GA4 for session-level channel behavior and for the custom AI grouping. When you compare them, reconcile on order count and revenue rather than sessions, because session definitions differ across the two systems. If Shopify shows AI-tagged orders that GA4 filed under Direct, that gap is your referrer-stripping problem, quantified.

AI is one percent of sessions. Why measure it at all?

Because the channel total is the wrong unit. One percent of sessions says nothing about whether the products you fixed last month now appear in answers and convert. Measure before and after on those products: appearances for a fixed set of shopper questions, AI-referred sessions to their pages, and orders from those sessions. Three separate numbers, reported separately. A small channel that moves on the products you touched is evidence. A channel total is not.

For why the session share undercounts AI's influence in the first place, see why the click count undersells AI traffic.

What to do first

Turn on the view: create the custom AI channel group in GA4, place it above Referral and Organic, and start collecting a clean baseline alongside the native channel.
Tag your feeds: add consistent UTM parameters to every product feed destination URL you syndicate to AI surfaces so app and agent clicks stop vanishing into Direct.
Label Google Shopping: enable Merchant Center auto-tagging for free listings so they separate from AI Mode inside Organic Search.
Reconcile monthly: pull AI-attributed orders from Shopify and compare them against GA4's AI channel to size your referrer-stripping gap.
Check whether you appear at all: if AI traffic is near zero, measurement is not the problem. The free AI readiness report shows how your products appear for a set of shopper questions before you tune the analytics.

Frequently asked questions

Does GA4 track ChatGPT traffic automatically now?

Yes. Since May 13, 2026, GA4's AI Assistant channel automatically classifies referrals from ChatGPT, Gemini, Copilot and other recognized assistants. It only captures visits that carry a referrer header, so app-based and agent browser visits without a referrer still land in Direct and need UTM tagging to recover.

Why does my AI traffic show up as Direct in GA4?

Because mobile apps, in-app browsers and agent browsers often strip the referrer. GA4 has no source to read, so it defaults those sessions to Direct. SEO practitioners estimate 20 to 40 percent of AI visits arrive this way. UTM parameters on links you control are the reliable fix.

Is Google AI Mode traffic counted as AI in GA4?

No. Google's channel definitions place AI Mode and AI Overviews in Organic Search and exclude them from AI Assistant. For retailers this is often the largest AI discovery source, so treating Organic as purely traditional search understates AI's real impact on your store.

Why do GA4 and Shopify analytics disagree on AI traffic?

The two systems use different attribution models, session definitions and data sources. Shopify sees the checkout and attributes orders with first-party data. GA4 uses a channel waterfall. Reconcile them on order count and revenue rather than sessions.

How much better does AI traffic convert?

Adobe observed that as of March 2026, AI-driven traffic to retail sites converted 42 percent more often than non-AI traffic, with 37 percent higher revenue per visit, as reported by Digital Commerce 360. We looked at that reversal in AI shoppers convert 42 percent better. Clean measurement is what lets you check whether your store sees the same pattern.

Once the view is clean, the next question is why the number is what it is. This is the work Paz does for lean ecommerce teams. A Paz expert finds the product pages and feed fields that keep you out of AI answers, you approve the fixes, and the platform publishes them and tracks the before and after in this same report. Talk to us.