ChatGPT is recommending WooCommerce stores. Perplexity is answering shopping questions with product links. Claude is helping users compare options before they buy. 63% of websites already receive AI-driven traffic — and ChatGPT accounts for roughly 50% of it. The problem isn’t that the traffic isn’t arriving. It’s that GA4 can’t tell you it came from an AI assistant at all.

This matters for your bottom line. If ChatGPT is sending you buyers and GA4 labels those sessions as “Direct” or buries them inside the catch-all “Referral” bucket, you cannot make informed decisions about your content strategy, your product pages, or which AI platforms deserve your attention. You’re optimising blind.

Two Kinds of AI Traffic — Only One Is Worth Measuring#

Before fixing your attribution, understand the difference between AI traffic types. Confusing them costs you accuracy in both directions.

AI crawler traffic comes from bots like GPTBot (OpenAI), PerplexityBot, and ClaudeBot. These index your content so AI assistants can answer questions about your products. According to Cloudflare, GPTBot’s crawler volume grew 305% from May 2024 to May 2025. PerplexityBot grew 157,490% in the same period. This is noise in your analytics — crawlers don’t buy. Filter them out.

AI referral clicks are real humans. A user asks ChatGPT for the best running shoes under £80 and clicks the link to your store. They arrive from chatgpt.com, perplexity.ai, or claude.ai carrying a human browser user agent. These sessions should convert at a meaningful rate — and they’re the ones you can’t see clearly right now.

You may be interested in: Your GA4 Audience Report Is Not Your Real Audience: How Consent Bias Skews Your Analytics

Why GA4 Gets This Wrong#

GA4 categorises generative AI referral traffic as standard referral traffic, because it arrives from another domain — chatgpt.com is a website, not a search engine. Search Engine Land confirmed in 2025 that GA4 has no native understanding of AI assistants as a distinct acquisition channel. Your traffic from Perplexity and your traffic from a random blog post sit in the same bucket.

It gets worse. When ChatGPT’s interface strips the referrer header — which happens in the app and certain response modes — GA4 receives the session with no referrer data. The classification: Direct. A converting buyer from ChatGPT becomes invisible, attributed to the channel that means nothing.

This isn’t a GA4 configuration error you can fix with a filter. It’s a structural gap in how GA4 was built — before AI assistants became referral channels at scale.

How to Make AI Traffic Visible in GA4#

The fix has two layers: a GA4 custom channel group for classification, and server-side UTM preservation for capture.

Step 1: Create a Custom Channel Group for AI Sources

In GA4, go to Admin → Data Display → Channel Groups. Create a new custom group and add rules that match sessions originating from AI assistant domains:

• Source contains chatgpt.com
• Source contains perplexity.ai
• Source contains claude.ai
• Source contains copilot.microsoft.com
• Source contains gemini.google.com

Name this channel group “AI Assistants” or “Generative AI”. GA4 will now surface these sessions as a distinct acquisition channel in your reports — separate from generic Referral and Direct.

Step 2: Filter Bot Crawlers from Your Reports

In GA4 Explore, add a filter excluding known bot user agents: GPTBot, PerplexityBot, ClaudeBot, GoogleOther. Alternatively, apply a server-side filter at your hosting or CDN layer to block these crawlers from generating events entirely. Don’t let them inflate your session counts or contaminate your funnel data.

Step 3: Use UTM Parameters in AI-Cited Content

If your content appears in AI assistant responses, you can influence how sessions are tagged. Pages with UTM parameters already embedded in their URLs will pass those parameters to GA4 when users click through. Structure your shareable URLs with consistent utm_source=chatgpt, utm_medium=ai-referral tags for pages you’re actively promoting in AI-visible content formats.

You may be interested in: When AI Agents Buy From Your WooCommerce Store, GA4 Sees Nothing

The Referrer Stripping Problem and Why Server-Side Tracking Solves It#

The ChatGPT app, browser extensions, and certain conversational interfaces strip the HTTP referrer header before passing users to your site. GA4’s client-side tag receives the session with no origin information. No referrer = Direct in GA4’s logic. You lose the attribution.

Server-side tracking changes the failure mode. Instead of relying on the browser to pass referrer data to GA4, a server-side setup captures the first-party context the moment the user lands — before any client-side script runs. When combined with persistent first-party cookies on your own subdomain, you retain attribution even when the referrer header has been stripped by the AI interface.

This is the same mechanism that recovers attribution lost to Safari’s ITP restrictions — the technical problem is identical. The AI referrer strip is just a new source of the same attribution gap your server-side setup should already be solving.

What This Looks Like in Practice#

A WooCommerce store selling physical products gets a content mention in a ChatGPT response about buying guides. Over 30 days, the store receives 340 AI-referred sessions. Of those:

• 180 arrive from chatgpt.com with referrer intact → classified as Referral in GA4
• 95 arrive from the ChatGPT app with no referrer → classified as Direct in GA4
• 65 arrive from perplexity.ai → classified as Referral in GA4

Without the custom channel group, the store owner sees 180 anonymous referrals mixed with all other referral traffic, 95 sessions added to a Direct total they assume is brand visits, and 65 more anonymous referrals. 340 potentially high-intent AI buyers are invisible as a channel. With the channel group set up and server-side UTM recovery in place, all 340 are captured, labelled, and attributable to conversion decisions.

The Transmute Engine™ Approach#

The Transmute Engine™ handles referrer recovery at the server level — meaning AI-referred sessions that lose their referrer header in transit are still captured with full attribution context via first-party cookies on your subdomain. The inPIPE™ plugin sends WooCommerce purchase events to the Transmute Engine server, where attribution data is preserved across the entire conversion path regardless of what the browser reports. GA4 receives the corrected, enriched event — not the stripped, incomplete version from a client-side tag.

For WooCommerce stores that care about understanding every acquisition channel — including emerging ones — this matters. AI referral traffic isn’t a niche edge case. It’s growing fast, converting well, and your current setup is probably missing most of it.