There are two different things to measure, and they answer different questions. First, the humans who arrive after an AI assistant recommended you — you find these in your analytics as referrals from domains like chatgpt.com, perplexity.ai, gemini.google.com and copilot.microsoft.com. Second, the AI crawlers themselves — bots like GPTBot, PerplexityBot and ClaudeBot fetching your pages, which show up in your server logs by user agent. Track the first to see AI-driven revenue; track the second to confirm engines can even reach you. You cannot improve what you are not measuring.
Every conversation about AI search skips the least glamorous, most important step: measurement. Merchants pour effort into getting cited, then have no idea whether it worked — because AI-driven visits are easy to miss in a standard analytics setup. If you cannot see the traffic, you cannot tell what is moving, and you end up optimizing blind. Here is how to make AI traffic visible, in two layers.
Two very different things to track
People conflate them, but they are not the same measurement — and you want both.
Humans referred by AI
Shoppers who clicked through to your store from an AI answer. This is your AI-driven demand, and it lands in analytics as referral traffic from the assistant domains.
AI crawlers fetching you
The bots that read your store to build answers in the first place. They never buy anything, but if they are not visiting, you cannot be cited — and they show up in server logs, not analytics.
Track the humans: AI referral traffic in GA4
When someone follows a link from an AI answer to your store, the assistant is usually recorded as the referrer. That makes AI-referred visits findable — once you know where to look.
Look for the AI source domains
In your analytics, filter referral or session-source reports for the assistant domains: chatgpt.com, perplexity.ai, gemini.google.com and copilot.microsoft.com are the main ones today. Anything from those is an AI-referred visit.
Group them into one channel
Rather than reading four separate rows, group those domains into a single custom channel or segment — call it something like AI Assistants. Now you can watch AI traffic as one trend line.
Follow them to conversions
Connect that segment to your store outcomes — add-to-cart, checkout, revenue. This is the number that tells you whether AI visibility is actually paying, not just flattering.
Watch which pages they land on
See which products and articles AI sends people to. Those are the pages engines already trust — they tell you what is working and what to build more of.
Track the crawlers: are AI bots even visiting?
Before any human can be referred, an engine has to have crawled you. That activity does not show in most analytics tools — because analytics runs on JavaScript that bots do not execute — but it is right there in your server logs.
Look through your log or hosting analytics for the crawler user agents: GPTBot and OAI-SearchBot from OpenAI, PerplexityBot from Perplexity, ClaudeBot from Anthropic, and Google-Extended from Google. Seeing them fetch your pages is proof the door is open. Not seeing them — or seeing them hit errors and blocks — is an early warning that no amount of content will get you cited, because nothing is being read.
Before you build dashboards, confirm the crawlers can actually reach you. Our free AI Visibility Audit checks AI-crawler access and whether your facts are server-rendered in seconds — the prerequisites that decide whether there is any AI traffic to measure in the first place.
AI referral tracking is imperfect. Some assistants strip or obscure the referrer, some traffic shows as direct, and voice or in-app answers may send no click at all. Treat AI-referral figures as a directional floor, not a precise count — the trend matters more than the exact number.
What to do with what you learn
Once you can see both layers, the strategy writes itself. If crawlers are not visiting, fix access first. If they visit but no humans are referred, your content is being read but not chosen — a schema, brand and answer-quality problem, which is what the AI-Ready Kits are built to fix. If humans arrive but do not convert, the AI part is working and the job moves to your product pages. Measurement turns a vague goal into a specific next move.
If you are still confirming whether engines can read you at all, start with how to check if your store is visible to AI; for the diagnostic on why a store gets read but not cited, see why your store is not showing up in ChatGPT.
A note on method. The referrer domains and crawler user agents listed here are the significant ones as of the last review, drawn from the AI platforms' own published crawler documentation. Both lists change as products launch and rename, so treat them as a starting set and review your own top referrers and logs periodically. AI referral measurement is inherently approximate. Last reviewed August 2026.
Track two things: humans referred by AI (analytics referrals from chatgpt.com, perplexity.ai, gemini.google.com and copilot.microsoft.com) and AI crawlers fetching you (server-log user agents like GPTBot and PerplexityBot). Confirm engines can reach you at all with the free AI Visibility Audit, then use the data to decide whether to fix access, content or conversion.