What counts as an 'AI-channel' sale in Shopify's reporting? Does it include a customer who just researched on ChatGPT but bought later through Google?
Shopify counts an AI-channel sale two different ways, and neither credits a ChatGPT research session that ends in a Google purchase. Sales channel attribution records where the order was placed. Marketing reports default to last non-direct click, which resets the referrer after 30 days (Shopify Help Center, September 2026). Google takes that sale.
Two systems, two definitions, one very common misreading
The reason this question has no single answer is that Shopify answers it twice, in two different reports, with two different meanings of the word channel.
1. Sales channel attribution. "Orders from AI channels display in your Shopify admin with channel or referrer attribution so that you know where the order originated from" (per Shopify Help Center, September 2026). Where direct checkout is active, on Google AI Mode and Gemini, Copilot and Meta, the order is genuinely placed in that channel and is recorded as such. This is a fact about where the transaction happened, not a model.
2. Marketing and referrer attribution. For any session that lands on your online store, including a ChatGPT in-app browser session, Shopify's marketing reports apply an attribution model. The default "Gives 100% of attribution credit to the last channel that the customer interacts with before making a purchase, excluding direct visits," and "If a visitor doesn't buy anything within 30 days of a session, then the first interaction referrer is reset to be whatever the next referrer is" (per Shopify Help Center, September 2026).
So the answer to the literal question is no, under the default model. Research on ChatGPT, purchase arriving from a Google click, and the sale is a Google sale in the reports your team looks at.
What changes the answer
| Model | Who gets the ChatGPT-research, Google-purchase sale |
|---|---|
| Last non-direct click (default) | |
| First click | ChatGPT, provided the ChatGPT session was the first interaction and the purchase happened within the 30-day window |
| Linear | Split across every click in the journey, including both |
Shopify lets you compare models in the marketing reports. Switching the default is not the right move for most teams, because every historical target was set against it. Running first click as a secondary view, purely to size the discovery contribution, is the move.
The leakage nobody adjusts for
Three separate effects push AI's measured contribution down, and all three are structural rather than fixable.
- Untagged sessions. Not every outbound AI link carries a tag, and untagged sessions land in Direct. The independent Shopify study we cite elsewhere in this thread estimates true ChatGPT volume at "1.1x to 1.6x" reported figures on that basis (per Ethercycle, August 2026).
- Cross-device. Research on a phone in a chat app, purchase on a laptop, and there is no session chain at all.
- The 30-day reset. Considered purchases with long cycles, which is most of what a $20M to $100M brand sells above $150, routinely exceed it. The study's own framing is that AI referrals "undercount by nature" because "when chatbots recommend products but shoppers search later, credit goes to search."
The honest consequence: your AI-channel number is a floor, not an estimate. Treat it as the smallest true value, and never build a case on the gap you imagine above it.
How to actually size the research contribution without inventing a number
- Run first click as a comparison view in the marketing reports for one quarter and report both figures side by side, always labelled.
- Add a post-purchase survey question with a named option for AI assistants. It is self-reported and imperfect, and it is the only instrument that sees a cross-device journey at all.
- Watch landing-page mix rather than volume. Shopify reports "more than half of AI-referred sessions start on a product detail page, compared to about 20% for organic search" (per Shopify, May 2026). A rising share of deep landings from any referrer is a discovery signal that survives attribution noise.
- Do not reconcile Shopify against GA4 on this. They apply different models to different session definitions, and the reconciliation project will cost more than the insight.
The Deploi point of view
Our own position, from building on Shopify. Separate from the facts above.
- Our take: Report the AI-channel number as a floor and say so in the same sentence every time. The default last non-direct click model is correct for budget decisions and structurally blind to discovery, and both of those statements are true at once.
- What we tell clients to do first: write down which of the two definitions your dashboard is using before anyone argues about the number. Half the disagreements we get pulled into are two people reading different reports and assuming they are reading the same one.
- Where we disagree: The AI-attribution tool category sells a fix for this, usually in the form of a proprietary model that assigns credit the platform's own reporting will not. We would not buy that at this volume. When a channel sits near a tenth of a percent of sessions, the cost of modelling it exceeds anything the model could change about a decision.
- What this page adds: that Shopify answers this question twice with two different definitions of channel, what the 30-day reset does to a considered purchase, and the three leakage effects that make the reported figure a floor.
Reviewed by Martin Dejnicki, Director of SEO & AI Search. Facts verified 2026-09-14.
Where we worked this out
Our decision records