How to measure AI traffic to your Shopify store
If an AI assistant recommends your product and the shopper clicks through and buys, that's the most valuable traffic you'll ever get: pre-sold, high-intent, and growing. It's also nearly invisible in a default analytics setup — it lands in “Direct” or generic “Referral” buckets and nobody notices. Here's how to actually see it.
What AI traffic looks like in the wild
AI-sourced visits arrive with recognizable fingerprints — a referrer from the assistant's domain, or a UTM the surface appends:
| Assistant | Typical referrer / marker |
|---|---|
| ChatGPT | chatgpt.com |
| Perplexity | perplexity.ai |
| Gemini | gemini.google.com |
| Copilot | copilot.microsoft.com |
| Claude | claude.ai |
| Google AI surfaces | often utm_source=google variants with AI-specific mediums |
Caveat up front: this undercounts. Some assistants open links without a referrer, some users copy-paste URLs, and in-app browsers strip headers. Treat measured AI traffic as a floor, not a census — trends month-over-month matter more than absolute counts.
Method 1 — Free and manual: your analytics tool
In GA4 (or Plausible/Fathom/Matomo), build a filter or segment matching the referrer domains above. In GA4: Reports → Acquisition → Traffic acquisition → add a filter on “Session source” containing each domain, or build one exploration with a regex matching all of them. Save it and check monthly.
- Pros: free, uses what you have.
- Cons: manual setup per domain (and new AI surfaces appear constantly), sessions only — tying AI visits to revenue and specific products requires joining ecommerce events yourself, and GA4 sampling/consent gaps hit small stores hardest.
Method 2 — UTM discipline for the surfaces you control
Where you control the link (your llms.txt, product feeds, marketplace profiles, social bios), append UTMs so any assistant citing those URLs carries attribution: ?utm_source=ai&utm_medium=assistant variants. It costs nothing and rescues a slice of the referrer-less traffic from Method 1's blind spot.
Method 3 — Automatic, first-party, tied to revenue
The structural fix is measuring at the store itself, not in a third-party analytics silo: a web pixel that recognizes AI referrers and UTMs at the session level, follows the session through checkout, and attributes revenue per product — so you can answer the only question that matters: “which products earn money from AI recommendations, and did that grow after I improved them?”.
This is what Preferd's AI Traffic Analytics does, and the design goals are worth stating because they generalize to any tool you evaluate: zero configuration (no GA4 wiring), first-party (your store, your data), zero personal data collected, and consent-aware. Attribution without a privacy bill.
Whichever method: close the loop
Measurement is only useful next to action. The loop is: audit your product data (the nine signals) → fix the worst products → watch whether category answers and AI-sourced revenue move. If you improve twenty products in July, August's AI-traffic report is your proof — dated, repeatable, and in euros rather than vibes.