If you want to know whether all the AI-readiness work is paying off, you need to measure it, and standard analytics will not show you. Google Analytics is built to track human sessions with JavaScript, and most AI agents do not run JavaScript or behave like a human session. So they are largely invisible in the reports merchants actually look at. Here is how to see them.
Why your normal analytics misses agents
Tools like GA4 work by running a script in the visitor's browser that fires events. AI crawlers and retrieval bots usually fetch your pages server-side, without executing that script, so they never register as sessions. Even the ones that do render JavaScript often get filtered as bots. The result is a real and growing slice of traffic that your dashboard simply does not count.
This is not a flaw you can configure away in GA. It is a structural mismatch. To see agents you have to look at a layer below the browser script.
Server logs are where the agents actually show up
Every request to your store, human or machine, hits the server and can be logged. That log, not the analytics dashboard, is where AI agents are visible. You identify them by their user agent string.
The names worth watching in 2026:
- GPTBot and OAI-SearchBot, the OpenAI crawlers behind ChatGPT and its search.
- ChatGPT-User, which fires when a person's ChatGPT session fetches a page on their behalf.
- PerplexityBot and Perplexity-User, for Perplexity.
- ClaudeBot, for Claude.
- Google-Extended, Google's signal for AI training and Gemini.
- Applebot-Extended, Amazonbot, and Meta-ExternalAgent for the others.
The distinction between the plain crawler and the "-User" variants matters. GPTBot is broad crawling. ChatGPT-User means a real person, in a real conversation, caused that fetch. The second is much closer to intent, and a rise in it is the clearest sign your store is being surfaced in actual shopping conversations.
How to get at the logs on Shopify
This is the practical snag. Shopify does not hand you raw server logs the way a self-hosted site would, so you have a few realistic paths.
If your store or marketing site sits behind a platform that exposes request logs, such as a Vercel or Cloudflare layer in front of a headless or hybrid setup, you can read agent hits there directly and even build a simple dashboard from them.
For the Shopify storefront itself, the cleaner route is an app that records agent requests to your store's AgentReady surfaces, the llms.txt, the structured data endpoints, and any feed, and reports them back to you. That gives you the agent view without needing raw Shopify logs, which is exactly the visibility most merchants are missing today.
A lightweight cross-check anyone can do: open Search Console and look at crawl stats, and watch for the AI user agents there over time. It is coarse, but it confirms the trend.
What the numbers actually tell you
Once you can see agent traffic, read it for three things.
Trend over time. Is agent traffic to your store growing? If you have been improving structured data and readiness, you want to see the crawlers and especially the "-User" hits trending up. That is the readiness work showing results.
Which agents. The mix tells you where you are getting picked up. Heavy GPTBot with little ChatGPT-User means you are being crawled but not yet surfaced in conversations. Rising ChatGPT-User means you are showing up in answers.
What they fetch. If agents are pulling your llms.txt and your product pages, your map is working. If they are only hitting the homepage and bouncing, your internal structure may not be guiding them to the catalog.
What to do with it
Measurement is only useful if it changes something. A few moves the data tends to prompt:
If crawler hits are high but "-User" hits are low, the gap is usually relevance and clarity. Tighten your structured data and descriptions so an assistant is confident enough to recommend you, not just index you.
If certain agents are absent entirely, check robots.txt. The most common cause of zero traffic from a given assistant is accidentally blocking its crawler.
If agents fetch the homepage and little else, improve internal linking and your llms.txt so the path to your products is obvious.
The honest state of this
Agent-traffic measurement is younger and rougher than human analytics. The user-agent strings shift, attribution from an agent answer to an eventual sale is still hard, and no tool does this as cleanly as GA4 does for human sessions. But the direction is clear, and a rough measurement of a real and growing channel beats a precise measurement of nothing. If you are investing in being readable to AI, you should at least be able to watch the agents arrive.
Surfacing this view is part of why we built the agent-traffic reporting in AgentReady, which records hits to your AgentReady surfaces so you can see which assistants are reading your store. However you get at it, start by looking below the browser script, because that is where the agents have been all along.