How do I track AI traffic to my Shopify store?

Two layers. First, referral traffic: in GA4, look for referrals from chatgpt.com, perplexity.ai, gemini.google.com, and copilot.microsoft.com, and build an exploration or channel group that isolates them, because by default they get scattered across Referral and Direct. Second, and more important, visibility: referral clicks only capture shoppers who clicked a link, and many AI answers name you without a clickable citation, so you also need to track how often the engines mention your products for your target prompts. Do both. Referrals tell you who clicked, prompt tracking tells you whether you are in the answer at all.

Why is AI traffic not showing up in my analytics?

Three reasons. Some AI surfaces send visitors without a referrer header, so the session lands in Direct instead of Referral, undercounting the true number. Some answers cite you but the shopper reads the answer and never clicks, so there is no session to measure at all. And some traffic arrives days later when the shopper returns via a branded Google search or types your URL, which last-click attribution credits to search or direct, not to the AI answer that actually started the journey. The traffic is real, the default reports just hide it.

What referrers do ChatGPT and Perplexity use?

The common ones to watch in GA4 are chatgpt.com (and openai.com), perplexity.ai, gemini.google.com, copilot.microsoft.com and bing.com for Copilot, and gemini or google referrers for AI Overviews, which are harder to separate from normal Google. Build a custom channel group or a regex filter on these hostnames so they stop hiding inside Referral and Direct. Expect the list to keep changing as products rename and merge, so treat it as a living filter, not a one-time setup.

Can Google Analytics track ChatGPT referrals?

Partly. GA4 will record a session as coming from chatgpt.com when the click carries that referrer, and you can isolate those in an exploration or a custom channel group. What GA4 cannot do is capture answers where you were cited but not clicked, or where the referrer was stripped, or where the AI answer influenced a purchase that closed later through another channel. So GA4 is necessary but not sufficient. Pair it with prompt-level visibility tracking to see the full picture.

How do I set up a GA4 channel group for AI traffic?

In GA4 admin, create a custom channel group, add a channel called AI Assistants, and define it with a condition matching source or referrer against a regex of the AI hostnames (chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, and so on). Once saved, that channel appears in your acquisition reports so you can watch AI-sourced sessions, conversion rate, and revenue over time. It will still undercount, for the reasons above, but it turns invisible traffic into a trend you can act on.

Does AI traffic convert better than regular search?

Often yes, because the shopper arrived after an assistant already answered their research question and effectively pre-qualified the choice, so they land closer to a decision. But be careful reading conversion rate off undercounted traffic: if half your AI sessions are misattributed to Direct, the AI channel you can see is a biased sample. Track the trend rather than the absolute rate, and lean on prompt visibility to judge whether you are winning the recommendation, which is the leading indicator that predicts the traffic.

Is UTM tagging useful for AI traffic?

Not really, because you do not control the links AI engines generate, so you cannot append UTM parameters to them the way you would to an email or ad. UTMs help for links you place yourself, like the link in your llms.txt or a page you syndicate. For organic AI citations, you are stuck with referrer analysis plus prompt tracking. That limitation is exactly why visibility monitoring matters: it measures the thing UTMs cannot.

What is the best way to measure AI visibility over time?

Pick the set of prompts your shoppers actually ask, run them across the major engines on a schedule, and record for each one whether your products are named, whether a competitor is named, and whether the mention links back to you. Chart share of voice over time and tie it to the changes you ship. Naridon does this across five engines and connects it to the fixes it applies, so you can see visibility rise as the gaps close rather than guessing from noisy referral data.

How to Track AI Traffic to Your Shopify Store (2026)

Want more qualified Shopify traffic from AI search?

Run a free Naridon scan to see which prompts, products, and AI engines can send more ready-to-buy visitors.

Start free scan

Right now, an AI assistant is answering a shopping question in your category and, sometimes, sending the shopper to a store. The uncomfortable part for most merchants is that when you open your analytics to see how much of this is happening, the number looks tiny or missing. It is not that the traffic is not there. It is that the default reports put it in the wrong bucket, and a large share of AI influence never shows up as a click at all.

This post is about seeing it clearly. There are two layers to track, referral clicks and visibility, and confusing them is why so many merchants either dismiss AI as hype or overpay for a tool that measures the wrong thing.

Layer one: referral traffic, and why it hides

When a shopper clicks a citation in an AI answer, the resulting visit can carry a referrer like chatgpt.com or perplexity.ai. In principle GA4 records that, and you can see it. In practice, three things bury it.

Missing referrers. Some AI surfaces send the visitor without a referrer header. With no referrer, GA4 files the session under Direct, the same bucket as someone typing your URL. Your real AI traffic is therefore split between a visible Referral slice and an invisible Direct slice.

Scattered sources. Even the referrals that do arrive are spread across several hostnames (chatgpt.com, openai.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, bing.com) and none of them roll up into a tidy AI channel unless you build one. Left alone, they look like noise.

Delayed conversions. Many shoppers read an AI answer, remember your brand, and come back later through a branded search or a direct visit. Last-click attribution then credits Google or Direct, and the AI answer that started the journey gets nothing.

The fix: build an AI Assistants channel in GA4

You cannot un-hide the missing-referrer traffic, but you can stop the visible portion from scattering. In GA4 admin, create a custom channel group and add a channel, call it AI Assistants, defined by a condition that matches the session source or referrer against a regex of the AI hostnames. Once saved, AI-sourced sessions, their conversion rate, and their revenue show up as their own line in your acquisition reports. It still undercounts, but a consistent undercount is a usable trend. Watch the direction, not the absolute.

One caveat: keep the hostname list current. These products rename and merge constantly, so treat the regex as a living filter you revisit, not a one-time setup.

Layer two: visibility, the thing referrals cannot measure

Here is the point most analytics guides miss. A huge share of AI influence never produces a click. The assistant names your product, states your price, summarizes your reviews, and the shopper acts on that without ever visiting a citation. No session, no referrer, nothing in GA4. If you only measure clicks, you are blind to the most important question: are you even in the answer?

That question is measurable, just not through web analytics. You measure it by tracking prompts. Take the questions your shoppers actually ask, run them across the engines on a schedule, and record for each whether your products are named, whether a competitor is named, and whether there is a link back to you. That is your share of voice, and it is the leading indicator: visibility rises first, and referral and branded-search traffic follow. This is the core of generative engine optimization measurement.

Why UTMs will not save you

The instinct is to tag links so attribution is clean. It does not work here, because you do not author the links AI engines generate, so you cannot append parameters to them. UTMs remain useful for links you place yourself, the link in your llms.txt, a syndicated page, an email. For organic AI citations you are limited to referrer analysis plus prompt tracking. That limitation is the whole reason visibility monitoring exists.

Putting the two layers together

Here is how the signals line up, and what each one can and cannot tell you.

Signal	What it measures	Blind spot	Where to see it
Referral sessions	Shoppers who clicked an AI citation	Missing-referrer and no-click influence	GA4 AI Assistants channel
Direct traffic lift	Rough proxy for stripped referrers	Cannot separate from real direct	GA4 Direct trend
Branded search lift	Delayed effect of AI exposure	Attributed to search, not AI	Search Console, GA4
Prompt share of voice	Whether you are in the answer at all	Not a direct revenue number	Visibility tracker

Read together, they triangulate the truth: prompt visibility tells you if you are winning the recommendation, the referral channel tells you the clicks it produces, and the direct and branded-search lifts confirm the influence the click data misses.

How Naridon measures it

Install Naridon free from the Shopify App Store, free forever at $0 with 150 credits a month, and it tracks your visibility across five engines, ChatGPT, Perplexity, Google AI Overviews, Gemini, and Copilot, for the prompts your shoppers ask. Because it also applies the fixes, it ties the two layers together: you can watch share of voice climb as gaps close, and correlate that with the referral and branded lift in your own analytics. Paid plans start at $49/mo with a 7-day trial. See the pricing page. For the wider method, the complete guide to GEO for Shopify covers the fixes that move the number.

AI traffic is not missing, it is misfiled and partly invisible. Build the GA4 channel to catch the clicks you can, accept that the direct and branded lifts hide more, and track prompt visibility to measure the influence that never becomes a click. That combination is how you actually see what AI is doing for your store.

Frequently asked

How do I track AI traffic to my Shopify store?: Two layers. First, referral traffic: in GA4, look for referrals from chatgpt.com, perplexity.ai, gemini.google.com, and copilot.microsoft.com, and build an exploration or channel group that isolates them, because by default they get scattered across Referral and Direct. Second, and more important, visibility: referral clicks only capture shoppers who clicked a link, and many AI answers name you without a clickable citation, so you also need to track how often the engines mention your products for your target prompts. Do both. Referrals tell you who clicked, prompt tracking tells you whether you are in the answer at all.
Why is AI traffic not showing up in my analytics?: Three reasons. Some AI surfaces send visitors without a referrer header, so the session lands in Direct instead of Referral, undercounting the true number. Some answers cite you but the shopper reads the answer and never clicks, so there is no session to measure at all. And some traffic arrives days later when the shopper returns via a branded Google search or types your URL, which last-click attribution credits to search or direct, not to the AI answer that actually started the journey. The traffic is real, the default reports just hide it.
What referrers do ChatGPT and Perplexity use?: The common ones to watch in GA4 are chatgpt.com (and openai.com), perplexity.ai, gemini.google.com, copilot.microsoft.com and bing.com for Copilot, and gemini or google referrers for AI Overviews, which are harder to separate from normal Google. Build a custom channel group or a regex filter on these hostnames so they stop hiding inside Referral and Direct. Expect the list to keep changing as products rename and merge, so treat it as a living filter, not a one-time setup.
Can Google Analytics track ChatGPT referrals?: Partly. GA4 will record a session as coming from chatgpt.com when the click carries that referrer, and you can isolate those in an exploration or a custom channel group. What GA4 cannot do is capture answers where you were cited but not clicked, or where the referrer was stripped, or where the AI answer influenced a purchase that closed later through another channel. So GA4 is necessary but not sufficient. Pair it with prompt-level visibility tracking to see the full picture.
How do I set up a GA4 channel group for AI traffic?: In GA4 admin, create a custom channel group, add a channel called AI Assistants, and define it with a condition matching source or referrer against a regex of the AI hostnames (chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, and so on). Once saved, that channel appears in your acquisition reports so you can watch AI-sourced sessions, conversion rate, and revenue over time. It will still undercount, for the reasons above, but it turns invisible traffic into a trend you can act on.
Does AI traffic convert better than regular search?: Often yes, because the shopper arrived after an assistant already answered their research question and effectively pre-qualified the choice, so they land closer to a decision. But be careful reading conversion rate off undercounted traffic: if half your AI sessions are misattributed to Direct, the AI channel you can see is a biased sample. Track the trend rather than the absolute rate, and lean on prompt visibility to judge whether you are winning the recommendation, which is the leading indicator that predicts the traffic.
Is UTM tagging useful for AI traffic?: Not really, because you do not control the links AI engines generate, so you cannot append UTM parameters to them the way you would to an email or ad. UTMs help for links you place yourself, like the link in your llms.txt or a page you syndicate. For organic AI citations, you are stuck with referrer analysis plus prompt tracking. That limitation is exactly why visibility monitoring matters: it measures the thing UTMs cannot.
What is the best way to measure AI visibility over time?: Pick the set of prompts your shoppers actually ask, run them across the major engines on a schedule, and record for each one whether your products are named, whether a competitor is named, and whether the mention links back to you. Chart share of voice over time and tie it to the changes you ship. Naridon does this across five engines and connects it to the fixes it applies, so you can see visibility rise as the gaps close rather than guessing from noisy referral data.

Key concepts

Plain-language definitions of the terms in this guide.

Ready to rank for these conversations?

Join early adopters who are already capturing AI search traffic.

Start free trial