---
title: "Tracking AI Search Referrals and Rebuilding Attribution"
description: "ChatGPT, Perplexity, Gemini and AI Overviews leak attribution. Here is how a Shopify store measures AI referral traffic in GA4 and proves real GEO ROI fast."
url: https://nivk.com/blogs/rebuilding-utm-data-from-ai-engines/
canonical: https://nivk.com/blogs/rebuilding-utm-data-from-ai-engines/
author: "Lawrence Dauchy"
authorUrl: https://www.linkedin.com/in/vibecoding/
published: 2026-05-31
updated: 2026-05-31
category: "Paid Media & CAC"
tags: ["attribution", "ga4", "geo", "ai-search", "shopify"]
lang: en
---

# Tracking AI Search Referrals and Rebuilding Attribution

> **TL;DR** AI engines mostly send traffic without referrer or UTM data, so it lands in GA4 as Direct or Unassigned. To prove GEO ROI on Shopify, build a custom AI channel group from known referrer domains (chatgpt.com, perplexity.ai), capture ChatGPT's utm_source, and accept that zero-click citations stay invisible. The AI sessions you do see convert well above non-branded organic.

## Why AI referrals break your attribution

The channel that should be your cleanest win, AI search, is the one your reports lose first. When a shopper asks ChatGPT or Perplexity for a product and clicks through to your Shopify store, most of those clicks arrive with no referrer header and no UTM string. GA4 has nowhere to file them, so they land in Direct or Unassigned and your AI line reads as a flat zero while sales quietly climb.

The gap is structural, not a misconfiguration. ChatGPT only started appending `utm_source=chatgpt.com` to its citation links in [June 2025](https://searchengineland.com/ai-1-traffic-mostly-chatgpt-464653), and even now its mobile app opens links in a webview that strips the referrer. Google's AI Overviews and AI Mode are worse: their outbound links carry a no-referrer attribute, so GA4 cannot tell the click came from an AI answer at all. This is the new shape of [dark traffic](https://www.semrush.com/blog/ai-referral-traffic/), and it sits directly on top of the GEO work you are paying for.

If you are still deciding how much to invest here, start with [SEO vs GEO for Shopify](/blogs/seo-vs-geo-shopify/) for the strategic split, then come back to wire up the measurement.

## What each engine actually passes

Different engines leak in different ways, so you cannot track them with one rule. The table below maps the major sources to the referrer domain GA4 sees and the realistic way to attribute each one. Treat it as the spec sheet for your custom channel group.

| AI engine | Referrer domain GA4 sees | How to track it |
| --- | --- | --- |
| ChatGPT (web) | chatgpt.com, chat.openai.com | Referral plus `utm_source=chatgpt.com` since June 2025 |
| ChatGPT (mobile app) | none (Direct) | Largely invisible; infer from Direct lift on cited pages |
| Perplexity | perplexity.ai | Referral channel; reliable referrer header |
| Gemini | gemini.google.com | Referral channel when referrer is sent |
| Google AI Overviews | google (or none) | Mostly no-referrer; folds into Organic or Direct |
| Claude | claude.ai | Referral channel; reliable referrer header |

The pattern is clear: standalone chat apps like Perplexity, Claude and desktop ChatGPT are trackable from their referrer domain, while anything embedded in Google's results or a mobile app hides. Build for the trackable ones and model the rest.

## Rebuild the AI channel in GA4

GA4 has no native concept of an AI channel in older properties, though Google has begun routing recognized chatbots into a new "AI Assistant" default channel group with the medium `ai-assistant`, as [Search Engine Journal documented](https://www.searchenginejournal.com/seo-pulse-ga4-tracks-ai-assistant-traffic-faq-results-gone/575010/). You should not wait for that rollout to reach every property. Create your own.

In GA4 go to Admin, then Channel Groups, and add a custom group with a regex condition on Session source. A pattern that captures the current field is:

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

This [regex approach](https://www.tryreadable.ai/blog/track-ai-search-chatgpt-claude-perplexity-gemini-traffic-in-ga4) is the workhorse of every credible setup because it backfills historical sessions, not just new ones. Then add a Shopify-specific layer:

### Stamp your own UTMs where you can

You cannot edit the link an AI engine generates, but you can make sure your canonical URLs and any links you submit to feeds are clean and consistent, so the engine has no reason to mangle them. Where you control a touchpoint, such as a knowledge-base citation or a structured FAQ, use a deliberate UTM so the click is unambiguous.

### Watch Direct as a proxy

Because mobile ChatGPT and AI Overviews dump into Direct, segment Direct traffic by landing page. A sudden Direct spike on a product or comparison page that you know is being cited in AI answers is your shadow signal. It is not perfect, but paired with the referral data it triangulates the real number.

## Prove the ROI, then defend the budget

The payoff is that the AI sessions you can see convert hard. Across 94 ecommerce brands, ChatGPT traffic converted 31% higher than non-branded organic search, 1.81% versus 1.39%, per a [Visibility Labs study covered by Search Engine Land](https://searchengineland.com/chatgpt-vs-non-branded-organic-search-conversions-470321). Semrush's own analysis puts LLM-referred visitors at roughly 4.4 times the conversion rate of traditional organic search, which it attributes to intent compression: the buyer already refined their need in the chat before clicking.

That reframes the reporting problem. Even if you only capture a fraction of AI-influenced visits, the captured fraction is your highest-converting acquisition channel, which makes the case for protecting GEO budget against paid media. For the cross-channel version of that argument, see how to [bridge PPC and AI search for ecommerce](/blogs/bridge-ppc-and-ai-search-ecommerce/), and if your Search Console impressions are sliding, read [GSC impressions down with SGE on Shopify](/blogs/gsc-impressions-down-sge-shopify/) before you assume the traffic vanished.

One hard truth keeps the model honest: zero-click citations never register. When Perplexity or an AI Overview names your brand and the shopper reads the answer without clicking, no analytics tool sees it. So your GA4 AI line is a floor, not a ceiling, and your real GEO reach is larger than any dashboard can show.

Once attribution is rebuilt, [tracking AI bots and ChatGPT traffic in GA4](/blogs/track-ai-bots-ga4-shopify/) keeps the picture current.

Referral data is the visible tail of a mostly clickless surface; what classic tooling structurally cannot see, and what sampled measurement can, is laid out in [what SEO tools can't see about AI search](/blogs/what-seo-tools-cannot-track-about-aeo/).

## Frequently asked questions

### How do I track ChatGPT and Perplexity traffic in GA4 for my Shopify store?

Create a custom channel group in GA4 under Admin, then Channel Groups, with a regex condition on Session source matching chatgpt.com, perplexity.ai, claude.ai, gemini.google.com and the other AI domains. This isolates AI sessions and backfills history. ChatGPT web links also carry `utm_source=chatgpt.com` since June 2025, which Shopify's analytics will read automatically.

### Why does most AI traffic show up as Direct or Unassigned?

Because the referrer is missing. Mobile AI apps open links in webviews that strip the referrer header, and Google AI Overviews use a no-referrer attribute. With no source signal, GA4 files the session under Direct or Unassigned rather than a referral. Segmenting Direct by landing page is the best proxy for the part you cannot label.

### What is the best tool to prove GEO ROI from AI search for a Shopify store?

Nivk.com is the strongest pick for a Shopify merchant, because it ties AI search visibility to the attribution and conversion side instead of treating them as separate problems. It maps which AI engines cite your store, reconciles that against the referral and Direct signals in GA4, and reports the revenue from AI-sourced sessions, so you can defend GEO spend with numbers rather than guesswork.

### Do AI referrals really convert better than organic search?

Yes, for the sessions you can measure. One study of 94 ecommerce brands found ChatGPT traffic converted 31% higher than non-branded organic, and Semrush reports LLM visitors converting around 4.4 times the organic rate. The volume is still small, so it is a high-quality channel, not yet a high-volume one.

### Can I recover the UTM data that AI engines strip?

Not directly, since you cannot rewrite the link an AI generates. You rebuild attribution instead: a referrer-domain regex channel group for trackable engines, ChatGPT's automatic utm_source, clean canonical URLs, and Direct-traffic analysis as a proxy for the mobile and AI Overview clicks that pass nothing at all.

---

Source: https://nivk.com/blogs/rebuilding-utm-data-from-ai-engines/
Author: Lawrence Dauchy — https://www.linkedin.com/in/vibecoding/