AI Drove 0.1% of Revenue: What $274M in Shopify Order Data Shows

What ethercycle measured, and how

No surveys, no dashboards, no vendor decks. ethercycle pulled raw order-level data from 10 established Shopify stores covering July 2024 through June 2026, then asked what share of sessions and revenue AI referrals actually represented. That method choice is the whole story: order data is tied to money that arrived in a bank account, and platform-reported numbers are not. Everyone who believed AI traffic was already large believed it on the strength of the second kind of number.

The same gap has a name on every channel

The lazy reading of the study is “AI shopping is hype.” The useful reading is that a platform narrative just met order data and lost. That exact gap exists on channels far older than AI: what Meta claims it drove versus what your orders show, what email claims versus what shipped. We call the compounding version of it attribution debt: inflated claims keep steering budget toward whichever channel is best at claiming credit, quarter after quarter. AI referrals are simply the newest and most visible example, and the first one people have bothered to audit this rigorously. The older gaps are still unaudited and still costing money.

The nuance most people skip

Three details from the study deserve more attention than the headline. First, those tiny AI referrals converted at 2.56%, against 1.75% for Google and a 1.45% portfolio average: in ethercycle's words, better than any channel they measured. Second, AI assistant sessions grew about 6x year over year, so a rounding error today is a real channel in a few compounding quarters. Third, the quiet winner was the Shop app, which outsold everything AI-related combined by roughly 35x ($4.8M against $137K). Low-volume, high-quality channels are precisely what a last-click dashboard is structurally incapable of surfacing, which is why the channels worth scaling are so often the ones your reporting renders invisible. That blindness is the same one that hides agent-placed orders from your pixel entirely.

Run the same audit on your own store

You do not need 10 stores and 94 million sessions to apply the method. Take last month's total revenue from your store admin. Then add up what every ad platform claims it drove in the same period. If the platform total exceeds your actual revenue, and for most multi-channel brands it does, the overshoot is your attribution debt, and it is deciding your budget right now. That reconciliation takes fifteen minutes. Going from “the numbers disagree” to “here is what each channel actually caused” is what a causal read on your GA4 export does, the same ask-the-orders principle, taken to channel level.

Measure it, then decide

The method behind the read: causal attribution, incrementality testing, and marketing mix modeling, priced at €99 one-time or €299/mo.

FAQ