How to track traffic and sales from ChatGPT, Gemini, Perplexity and Copilot

When a shopper clicks a link inside an AI answer, the browser usually passes the referring domain. Look for these in the session source report:

• chatgpt.com for ChatGPT. OpenAI also appends utm_source=chatgpt.com to links from ChatGPT search results, which gives you a second, more durable signal when the referrer is dropped. Source: the OpenAI publishers and developers FAQ.
• perplexity.ai for Perplexity answers and product cards.
• copilot.microsoft.com for Microsoft Copilot.
• gemini.google.com for the Gemini app.

Google AI Overviews and AI Mode sit inside Google Search, so their clicks arrive as google.com organic traffic. Google reports them separately in Search Console, covered below.

Google Analytics 4 includes an AI Assistant default channel for traffic from sources like ChatGPT, Gemini, Copilot or Grok. Google states that it excludes Google AI Overviews and AI Mode, which remain inside Organic Search. See the GA4 default channel group documentation. Two things push AI traffic out of that channel:

• Direct. Some assistants open links in an in-app browser or strip the referrer. The session arrives with no source and GA4 files it under Direct.
• Unassigned. A link with campaign parameters that match no channel rule, for example a custom utm_medium, lands in Unassigned. Standardize the medium on any AI links you control.

To track it reliably, open the Traffic acquisition report, switch the dimension to session source / medium and filter on the four domains above. For one line per assistant, build a custom channel group with a rule per domain and leave the default group as it is so historical reports stay comparable.

Shopify stores with agentic storefronts active get a report under Sales channels, then Agentic. For each AI channel it shows sales, orders, online store sessions and conversion, plus a total across AI channels, and orders carry channel or referrer attribution. Source: the Shopify agentic storefronts documentation. Two details matter when you read it:

• Shop is counted as an agentic storefront. Sessions, referrals and sales from the Shop app are included in the reporting. Read the ChatGPT, Copilot and Google lines on their own when you want the AI assistant number, and read Shop as its own channel.
• Sales aggregate two paths. Shopify adds referral-based sales, finished on your online store checkout, to direct checkout sales completed inside the AI channel. GA4 only sees the first path, so the Shopify number normally runs higher than the GA4 AI Assistant number for the same period. Write that difference down once and read both reports against it.

Analytics records the clicks that reach your site. Two platform reports show the step before the click:

• The Search Console generative AI performance report shows impressions and clicks from generative AI features on Google Search, by page, device and country. Google rolled it out to all sites worldwide on August 31, 2026, and it is the place to read AI Overviews and AI Mode.
• The Bing Webmaster Tools AI Performance report shows how often your pages are cited across Microsoft Copilot and AI summaries in Bing, which pages are cited most and the phrases Copilot used to retrieve them. Bing notes that the count is separate from position or role in the answer.

Read these next to the GA4 and Shopify numbers. A shopper who reads your product in a ChatGPT answer and searches your brand an hour later shows up as branded search, one of the reasons the attributed number is a floor.

Every order tied to an AI referrer is a real order. The orders without one are unknown, and several mechanics push AI-influenced purchases there:

• The assistant opened the link without a referrer, so the session is Direct.
• The shopper read the answer on a phone and bought on a laptop two days later.
• Consent settings limited the session data.
• The shopper read the answer, remembered the brand and searched for it later.

So treat each report as a floor on AI influence and the trend as the signal. The AI shopping attribution entry covers the measurement rules in detail.

A rise in AI-referred sessions after a round of product information fixes is directional evidence. It gets stronger with a few habits:

1. Date the baseline. Record GA4 AI Assistant sessions and orders, the Shopify agentic report and the two citation reports for the same window before any change ships.
2. Track the questions shoppers ask. Run a fixed set across ChatGPT, Google AI Mode, Gemini, Perplexity and Copilot on a schedule, and record whether your products appear and how.
3. Change the product information behind a specific gap. A missing fit attribute, a product page without an answer about compatibility, a feed disapproval.
4. Read the same reports over the following weeks, next to seasonality and promotions.

This is how Paz reports AI visibility work: appearances per assistant over time, next to the AI-referred traffic and orders in your analytics. Our experts set the baseline on day one, fix the product information behind each gap within the guidelines you approve, bring exceptions to you and show you what happened next. See the AI visibility service and the measurement methodology.