Are Omnilux's '3.2% of revenue from AI channels' and Cozy Earth's '20x YoY AI-channel growth' numbers typical, or are those the exceptions Shopify chose to showcase?
Shopify's published Omnilux and Cozy Earth AI-channel figures are real, selected, and not a base rate. Omnilux saw AI channels drive 3.2% of total revenue in March 2026, and Cozy Earth reported AI-channel revenue up 20x year over year (Shopify, June 2026). An independent 10-store Shopify sample measured AI referrals at 0.10% of revenue (Ethercycle, August 2026).
Both quotes, verbatim, with what each one is
From Shopify's Spring '26 Edition merchant post, published 17 June 2026: "In March, red light therapy mask brand Omnilux saw AI channels drive 3.2% of total revenue," and "according to luxury bedding brand Cozy Earth, overall revenue from AI channels was up 20x YoY."
Notice the difference in how those two sentences are constructed. The Omnilux figure is stated by Shopify. The Cozy Earth figure is prefaced with "according to," which is Shopify attributing the number to the merchant rather than publishing it as platform data. That distinction is in Shopify's own sentence and it is worth carrying forward.
The three reasons neither figure is a base rate
1. They are examples in a product announcement. Shopify published them in the edition post that launched agentic storefronts. Selection is the entire purpose of the artefact, and no reasonable reading of a product launch treats its named merchants as a random sample. That is not a criticism of Shopify. It is what a launch post is.
2. A multiple with no base is not a size. "Up 20x" tells you nothing about the destination. If AI channels were 0.05% of revenue a year earlier, a 20x lands at 1%. If they were 0.005%, a 20x lands at a tenth of a percent. Both are consistent with the quoted sentence and they are two very different businesses. Shopify's platform-level disclosures have the same property, growing 8x, 13x and 3x across two quarters without ever publishing the denominator.
3. The category is not a random draw. Omnilux sells a red light therapy mask. That is a high-consideration, specification-heavy, comparison-shopped product, which is precisely where Shopify itself reports the largest AI advantage: AI-referred sessions "outperform organic SEO in 23 of 25 merchant categories by an average of 56%," and more than half of AI-referred sessions land directly on a product detail page (per Shopify, May 2026). A basics apparel brand is not in the same measurement.
What a base rate would look like, for contrast
| Figure | What it is | Where it comes from |
|---|---|---|
| 3.2% of total revenue | One merchant, one month (March), specification-heavy category | Shopify, June 2026 |
| 20x year over year | One merchant, self-reported, base undisclosed | Shopify quoting Cozy Earth, June 2026 |
| 0.10% of revenue, 2026 year to date | 10 United States Shopify stores, 94M sessions, $274M revenue, last non-direct click | Ethercycle, August 2026 |
| Organic still exceeds all AI platforms combined | Platform-level bound on how large AI can be | Shopify, May 2026 |
The gap between 3.2% and 0.10% is roughly thirty-two times. That gap is the answer to the question. Both numbers can be true, of different merchants, in different categories, measured in different months, because the variance between brands in this category is currently far larger than the trend.
How to use a showcase number honestly
- As a ceiling. 3.2% is what a well-suited product in a well-suited category reached in one month in early 2026. Treat it as the top of the plausible range for a brand like that, not the middle of the range for a brand like yours.
- As a category test. If your product is specification-heavy, comparison-shopped and reads well as structured data, you are in the Omnilux half of the distribution. If it is impulse, fashion-led or brand-loyal, you are not, and no amount of feed work moves you there.
- Never as a forecast input. A board model that assumes 3% of revenue from AI channels next year is assuming the showcase case, in a category that may not be yours, off a base nobody published.
When NOT to bring these numbers into a meeting
- When the meeting is about budget. A selected example cannot size a spend.
- When you have your own admin data. Your figure is more relevant than either of theirs, at any magnitude.
- When the claim is being made about "brands like us." Ask which month, which base, and who measured it. Both public figures fail at least one of those three.
The Deploi point of view
Our own position, from building on Shopify. Separate from the facts above.
- Our take: Both figures are real and both are ceilings. We use the Omnilux number as the top of a plausible range for a specification-heavy product, and we refuse to use the Cozy Earth number at all, because a multiple whose base nobody published cannot inform a decision.
- What we tell clients to do first: pull your own AI channel sessions and revenue from the admin before anyone quotes a case study in a planning meeting. It takes ten minutes and it ends the argument.
- Where we disagree: The standard treatment of platform case studies is to discount them by some vague factor and use them anyway. We think the correct discount is unknowable here, because the missing piece is not the magnitude but the base, and a discount on an unknown base is still unknown. Use them as existence proofs that the ceiling is above zero, and use your own data for everything else.
- What this page adds: the exact wording of both Shopify sentences, including the "according to" that marks one of them as merchant-reported, the arithmetic that makes a 20x uninformative, and the roughly thirty-two-fold gap between the showcase figure and the only independent Shopify measurement.
Reviewed by Martin Dejnicki, Director of SEO & AI Search. Facts verified 2026-09-14.
Where we worked this out
Our decision records