Track ChatGPT and AI Referrals to Your Shopify Store

AI referral analytics separates visits with recognizable AI-source data from the larger set analytics can't reliably attribute.

For a Shopify store, start with Google Analytics 4 (GA4) and Shopify's Sessions by referrer report. Both reports show only the AI visits they can identify.

I've run Shopify stores, and I learned early that a traffic report is useful only when you know what it leaves out. You can run the reports below again without drawing conclusions from a few visits.

Key takeaways

What can AI referral analytics actually show?

AI referral analytics shows visits with usable source data, while the full effect of AI discovery remains unknown. A referrer or tagged source can connect a session to an AI assistant. Copied links, stripped referrers, and Google AI features may place related visits elsewhere.

Think about two shoppers who open the same product page. One clicks a ChatGPT link that passes a known referrer. The other copies the product URL into a new tab. Your reports may call the first an AI referral and the second Direct, though AI influenced both visits.

That gap is why I label the result "observed AI referrals." The label reminds you that the count is incomplete and doesn't measure every touchpoint. Our guide to what GEO means covers the visibility work that can happen before a measured visit.

Use the same questions for each free view:

‍

Start with the AI Assistant channel in GA4

GA4's default AI Assistant channel is the fastest first view of sessions from known AI-assistant referrers. Google's default channel documentation explains the rule. GA4 sets the medium to ai-assistant and the campaign to (ai-assistant) when a referrer matches its list.

Open Reports > Acquisition > Traffic acquisition. Set the main dimension to Session default channel group and find AI Assistant. Add Landing page as a comparison or secondary dimension. The landing page tells you where the session began, which gives the later review a page to inspect.

I use the Google-maintained native row as a baseline. Treat it as incomplete because visits outside Google's source list stay elsewhere, and Google can change that list. A custom view can then group source patterns that need a closer look.

Google places its own AI Overviews and AI Mode traffic in Organic Search. Keep that boundary intact because moving those visits into AI Assistant would mix two definitions and make later comparisons useless.

The Two-Tool AI Referral Check

Run the native GA4 view, custom fallback, Shopify report, and decision record in that order. Each check produces a saved output for the next one:

1. Save the GA4 native view: Keep the default baseline and landing page.
2. Add a custom fallback channel: Capture documented source patterns in another view.
3. Compare the Shopify report: Look for the same referrer and date.
4. Set the decision rule: Act only on a pattern that repeats.

After the checks, you can compare GA4's known sources, your custom matches, and Shopify's referrers before you choose what to inspect.

1. Save the GA4 native view

Save one GA4 view with a fixed date range, AI Assistant row, and landing-page dimension. This becomes the baseline for the later checks. The worksheet below holds the dates, source, sessions, and leading landing page.

Use a full past period because GA4 may still process today's events after you open the report. Keep the same duration when you repeat the view. A calendar month is easy to repeat, but judge the evidence from the traffic you receive.

I prefer a small worksheet over a dashboard at this stage because the worksheet forces you to write down what changed. Give it these columns:

• Review date: The day you saved the report.
• Report period: The exact start and end dates.
• Source: The visible referrer or source value.
• Landing page: The first page recorded for the session.
• Observed sessions: The report's session count.
• Orders shown elsewhere: Copy the value from the relevant sales report.
• Notes: Stock, page, campaign, or tracking changes during the period.

A single-day spike belongs in the notes. Wait for a repeat before adding it to a page-change plan. Your next check should use the same dimensions so any difference has a clear reference point.

2. Add a custom AI fallback channel

A custom channel group gives you a rule-based view of AI-source patterns you choose to include. Google provides an example instead of a permanent master list. Copy the current rule from its custom channel guide, then keep a change log beside your worksheet.

In Admin, open Data display > Channel groups. Copy an existing group and add a channel named AI assistants. Choose Source, select matches regex, and enter Google's example:

^.*ai|.*\.openai.*|.*chatgpt.*|.*gemini.*|.*gpt.*|.*copilot.*|.*perplexity.*|.*google.*bard.*|.*bard.*google.*|.*bard.*|.*.*gemini.*google.*$

Move the new channel above Referral because GA4 uses the first matching channel. A standard property can hold 2 custom groups besides the preset group. Update the same named group and log each regex or order change.

The rule groups source data that reaches GA4 and matches its terms.

Google says the expression needs an update when assistant URLs change. I'd record the date, regex, and channel order after each edit so a changed rule doesn't look like a traffic trend. Keep the old value in the log, since it explains why an earlier report may group the same source differently.

3. Compare Shopify's Sessions by referrer report

Shopify's Sessions by referrer report gives you a no-GA4 view of direct, search, and referral sessions. In Shopify admin, open Analytics > Reports. Filter the list to Acquisition and select Sessions by referrer. The report covers the Online Store sales channel.

Match the report period to the GA4 record. Inspect the relevant referrer, then copy that source label and its session count into the worksheet. The landing page still comes from the GA4 view. If a label is new, record it and classify it only after you verify the source.

Shopify changed its session measurement from September 21 to 23, 2026. A range that crosses those dates may show the change in method. Mark the break in your worksheet, and compare periods on the same side when possible.

Use this report for another session count. Check sales and order values in Shopify's Total sales by referrer report, and use GA4 for events and conversions.

When the session counts differ, compare the date range and raw referrer first. Then check the custom channel order in GA4. Keep the totals separate because the tools group traffic under their own rules.

4. Decide what the report is allowed to tell you

Investigate when the same source and landing page appear again, and keep tracking when they appear only once. Review the same saved views next month. Keep sessions apart from orders, page edits, stock changes, ad spend, and other moving inputs.

Suppose the same product page appears with the same AI source in two reviews. Open the GA4 landing-page row and confirm the referrer. Then compare the visible title, description, price, stock, delivery, and return facts with the current offer.

Change one weak or inaccurate element so the next review tests a clear explanation. Our guide shows how to improve the repeated landing page from verified product facts instead of rewriting the whole catalog.

If the pair disappears, keep logging it. Use repetition rather than a universal session threshold because small stores receive different traffic levels. A repeat tells you what to inspect next.

Record the decision in one of three plain states:

1. Repeat and investigate: The same source and page appear again.
2. Changed input: A page, stock, campaign, or rule changed between periods.
3. Keep measuring: The signal didn't repeat or remains too thin to interpret.

Edit one inaccurate product fact or one unclear description passage, then log its date. Compare the same saved report after the next full period. That keeps the test readable even when AI referrals stay low.

What the two reports cannot prove

These reports show identifiable sessions, while order attribution and hidden discovery paths remain outside their scope. Keep these limits in view:

• Google AI features: GA4 records AI Overviews and AI Mode under Organic Search.
• Direct sessions: Many paths lead to Direct, so its AI share remains unknown.
• Crawler visits: Bot requests describe access rather than human referral sessions.
• Order attribution: A source label alone can't assign an order to an answer.

Our guide to LLM crawler access covers whether named crawlers can request your pages. That is separate from the human sessions measured here.

AI citation work can affect discovery even when analytics records a later visit under another source. A citation and a recorded session describe different parts of the path.

Shopify separates session reports from sales reports. GA4 results depend on the events and model set for your property. Chats that send no source data remain outside both reports.

I recommend keeping the GA4 and Shopify views separate because each groups traffic under its own rules. Log page edits beside the report instead of blending both tools into one "AI traffic" total.

FAQ

Does AI Assistant include AI Overviews or AI Mode?

GA4 puts non-ad AI Overviews and AI Mode clicks in Organic Search rather than AI Assistant. That classification means the AI Assistant row can't provide a separate Google AI Mode total.

What does utm_source=chatgpt.com mean?

It means the session URL carried chatgpt.com as its UTM source value. When that value appears in GA4's Session source or Shopify's raw UTM data, copy it into the worksheet's source field. Untagged links and purchase causation require separate evidence.

Can Shopify tell me which AI referrals turned into orders?

Use Sessions by referrer for visits and Total sales by referrer for sales and order amounts. Compare those reports as separate views because a referrer still doesn't prove what caused an order.