How Can Shopify Stores Track ChatGPT and Perplexity Referrals in GA4?

Short answer: Shopify stores can identify some ChatGPT and Perplexity visits in GA4 through Traffic acquisition, using Session source / medium. Search for source values such as chatgpt.com and perplexity.ai, then review landing pages and recorded shopping activity. These rows show identifiable sessions, not every AI-influenced visit, citation, or recommendation.

If GA4 is already installed, start with its acquisition reports rather than adding another tracking tool. The job is to establish which visits GA4 can identify, what those visitors did, and where the measurement stops.

Where do you find ChatGPT and Perplexity traffic in GA4?

Open Reports → Acquisition → Traffic acquisition and change the table’s primary dimension to Session source / medium. Search each platform’s source separately, without first restricting the report to the Referral channel.

1. Select the GA4 property connected to your Shopify store. Choose a date range containing recorded store activity.
2. Open Traffic acquisition. Navigation can vary in customized properties. If the report is missing, someone with Editor access or higher can restore it.
3. Change the primary dimension. Use the dropdown above the table to select Session source / medium, or Session source if you want source names alone.
4. Search for chatgpt.com. Note the source, medium, sessions, and engaged sessions that appear.
5. Clear the search and check perplexity.ai. Record the actual values rather than assuming both platforms have the same medium.

A source identifies where traffic came from. A medium describes how that traffic is categorized, such as referral. A channel group combines traffic into broader categories using rules. Searching source values across channels avoids overlooking an identifiable source because it was grouped somewhere unexpected.

Use session-level dimensions for this task. Google’s Traffic acquisition documentation distinguishes them from First user source / medium, which describes how a visitor was originally acquired. A returning customer might first discover your store through Google and later arrive through an AI answer. Those are different acquisition questions.

Which source names should you look for?

chatgpt.com and perplexity.ai are supported examples to investigate, not a permanent or exhaustive detection list. Inspect the source and medium actually recorded in your property.

• ChatGPT: OpenAI’s publisher FAQ states that ChatGPT search referral URLs include utm_source=chatgpt.com. That is a campaign parameter attached to a link. The statement covers ChatGPT search referrals, not every ChatGPT surface or click path.
• Perplexity: Trakkr’s Perplexity Traffic Index identifies perplexity.ai among session sources in its connected GA4 properties. This is an observed analytics example, not an official guarantee about every Perplexity browser or app visit.

Do not require the exact combination chatgpt.com / referral or perplexity.ai / referral before investigating a row. Source information can come from campaign parameters or referring URLs, and the medium may differ.

If an exact search returns nothing, broaden the table search to the platform name and inspect any matching values. A matching name is a candidate to examine, not automatic proof of its origin. Keep a small record of the source values you have identified in your own reporting.

How do you save a reusable AI-referral exploration?

Create a Free form exploration with session sources and landing pages as rows, sessions and engaged sessions as values, and a source filter for one platform. Duplicate the tab for the other platform so you can repeat the review without rebuilding it.

1. Go to Explore → Free form. Give the exploration a recognizable name, such as “AI referral sessions.”
2. Import dimensions. Add Session source, Session source / medium, and Landing page + query string.
3. Import metrics. Add Sessions and Engaged sessions.
4. Build the table. Put Session source / medium and Landing page + query string in Rows. Put Sessions and Engaged sessions in Values.
5. Apply a source filter. Filter Session source to exactly match a value identified in your acquisition report.
6. Duplicate the tab. Change the source filter to the other platform’s observed value.

GA4 combines multiple exploration filter clauses with AND. Adding one exact source filter for ChatGPT and another for Perplexity does not mean “either platform.” It asks a source value to match both, which can leave the table empty. Separate tabs are easier to maintain.

Landing page + query string identifies the path and query string of the first page viewed in a session. This helps you distinguish visits that start on a blog article from those starting on a collection or product page. Query parameters can split the same underlying page into multiple rows, so inspect those variations before treating them as different destinations.

What should you measure after identifying the sessions?

Review landing pages, engaged sessions, product-view events, purchases, and relevant key events. These show recorded behavior associated with identifiable sessions, not sales proven to have been caused by ChatGPT or Perplexity.

• Landing pages: Which articles, collections, or products were the entry points?
• Engaged sessions: How many sessions met GA4’s engagement criteria? Use this alongside session counts, not as a direct measure of customer satisfaction.
• Product views: Check recorded view_item events to see whether product-detail viewing occurred.
• Purchases: Review recorded purchase events and revenue where ecommerce tracking is working.
• Key events: Review important actions you have designated in GA4. “Key events” is GA4’s terminology for these actions, which merchants may still call conversions.

In Traffic acquisition, use the Event count selector to inspect a specific collected event where available, and the Key events selector to review a selected important action. Confirm that the events exist and are being collected correctly. Installing GA4 does not, by itself, establish that every Shopify interaction you want to measure is recorded.

For example, a source row with purchases supports the statement “GA4 recorded purchases in sessions attributed to this source.” It does not establish that the platform alone persuaded those customers to buy.

For the broader content-review routine, see What Should Shopify Merchants Track After Publishing a Blog Post?

Why might AI visits be missing or appear as Direct?

AI-origin visits may lack identifiable source information because browser policies, app handoffs, privacy controls, redirects, and platform tagging affect what reaches GA4. Missing rows do not prove that no AI-influenced visits occurred, but unexplained Direct traffic should not be relabeled as AI traffic.

Google defines (direct) / (none) as traffic without a clear referral source. That category is not a hidden ChatGPT or Perplexity report. Some visits may also go unrecorded when analytics collection is blocked or limited.

Session behavior matters too. GA4 keeps one traffic source per session. Encountering another source during an existing session does not automatically start a new session or replace its session-level source. One AI click therefore does not necessarily produce one new AI-attributed session.

For troubleshooting, add Page referrer to a separate diagnostic exploration. It records the preceding URL supplied through page_referrer, which may be another store page or an external domain. Browser referrer policies can limit it to an origin or omit it entirely. Page referrer is not interchangeable with Session source, and it does not reliably reveal a shopper’s prompt or the AI answer they saw.

Keep these measurement boundaries separate:

• Identifiable referral sessions: Recorded visits with usable acquisition information.
• Citations and mentions: Appearances of your store or content in an AI response, which may never produce a click.
• Impressions: Exposure to an answer or link, not measured by a GA4 source row.
• Assisted influence: A possible contribution to someone’s research or purchase journey, which one acquisition row cannot establish.

How should you review AI referrals regularly?

Save the exploration and review it monthly, or weekly if visit volume makes that useful. Keep date windows and filters consistent, record meaningful behavior, and note tracking changes before interpreting increases or decreases.

A simple review log can contain:

• Date range and source / medium.
• Sessions and engaged sessions.
• Leading landing pages.
• Recorded product views, purchases, and selected key events.
• Changes to analytics, consent settings, redirects, or event collection.

Small counts can move sharply without indicating a durable pattern. Use the report to choose pages worth reviewing, not to declare that an article has “won” AI search.

Our SEOBoss visibility workflow complements this review with connected Google Search Console data, including article impressions, clicks, click-through rate, and average position. Those signals support content review, but they do not identify ChatGPT or Perplexity sessions or replace GA4 attribution. How Do People Find Your Shopify Store? puts these visits within the wider discovery picture.

The useful outcome is a repeatable record of identifiable sources, entry pages, and on-site actions. Keep that record separate from citations and unmeasured influence, and you can use AI-referral reporting without asking it to prove more than it can.