8 min read
AI is ecommerce's fastest-growing channel. Your wholesale revenue can't see it.
The multipliers made the headlines. The base rate did not. Read together, they say something specific about wholesale that nobody has bothered to say yet.
The number Shopify left out
Ethercycle published a State of Ecommerce 2026 study covering 94 million sessions across ten Shopify stores and $274m in revenue. Three of their findings travelled well, largely because Shopify said broadly similar things on its Q2 earnings call.
- AI visitors landing on a product page complete checkout at 2.56x the site-wide rate (Ethercycle). Shopify told investors the figure was around 2.5x.
- 85% of AI-referred revenue came from first-time customers, against a 42% portfolio average (Ethercycle). Shopify described new-buyer orders from AI channels running at roughly twice other channels.
- AI referrals grew 6x year over year in Ethercycle's portfolio; Shopify reported roughly 3x over its own window.
The figure that did not make the earnings call is the one that gives the others their shape. In Ethercycle's data, AI accounts for roughly 0.1% of traffic. Credit to them for publishing it, because a multiple without a base rate is a decoration rather than a measurement.
Both facts hold at once. AI is a rounding error in volume and, simultaneously, the fastest-growing, highest-converting, most new-customer-heavy channel anyone has measured on Shopify. Neither observation cancels the other, and the interesting question is not whether the trend is real. It is who gets to participate in it.
Why AI referrals convert so well
The conversion advantage is not magic and it is not novelty. It is a mechanism, and the mechanism is worth stating precisely, because it is the mechanism — not the statistic — that determines whether the effect transfers to another kind of commerce.
A shopper describes a need to a model in ordinary language. The model interprets that need, compares options, forms an opinion, and hands over a single link. What arrives on the merchant's product page is not a browser. It is a buyer who has already been through discovery and consideration somewhere else, arriving at the end of the funnel with the decision substantially made.
In other words, the model performs the pre-qualification that a storefront normally has to perform itself. Category pages, filters, comparison tables, review widgets — the entire apparatus of consideration is collapsed into one step that happens off-site. Of course the resulting session converts better than average. It is not the same kind of session.
That collapse works under four conditions, all of which happen to be true of most direct-to-consumer purchases: one person decides, one session is enough, the price is visible on a public page, and the purchase is low-consideration enough that a model's judgement is good enough to act on.
Now take those four conditions into wholesale
Every one of them fails, and they fail structurally rather than incidentally.
One decision maker
A wholesale purchase is rarely one person. There is a buyer, often an approver above them, frequently finance, sometimes an account manager on the supplier side and a category team on the buyer side. A model's recommendation does not survive contact with an approval chain, because the person receiving the recommendation is not the person authorised to act on it.
One session
Reorder cycles run in weeks and months. The gap between deciding to restock and placing the order is filled with stock counts, forecasts, budget windows and, very often, a phone call. There is no single session for a model to compress.
One visible price
Wholesale pricing is negotiated and customer-specific by design. Tiers, volume breaks, per-SKU overrides for key accounts — the entire point of the structure is that the price depends on who is asking. A model cannot surface a price that is, correctly, not printed anywhere public.
Discovery
A wholesale buyer usually knows the supplier and the SKU before the session starts. What they are actually checking is terms: minimum order quantity, current stock, lead time, payment terms, whether last quarter's price still stands. That is verification work, not discovery, and it is the part of the funnel where AI referral currently contributes least.
And the hard one: the catalogue is gated
A crawler cannot read a login-gated catalogue, and a model cannot recommend a product or a price it was never permitted to see. Every reasonably built wholesale portal hides its catalogue and its pricing behind authentication, because leaking negotiated trade prices to the public internet is precisely the failure mode the portal exists to prevent. The privacy that makes B2B pricing work is the same property that makes it invisible to the systems now driving the fastest-growing referral channel in ecommerce.
The implication, stated plainly
The fastest-growing acquisition channel in ecommerce is structurally invisible to the portion of revenue that sits behind a login.
If a merchant does 40% of revenue through wholesale, that 40% is not participating in this trend at all. The 2.5x and the 85% and the 6x are being earned entirely by the direct-to-consumer half of the business, and averaging them across the whole company produces a number that describes nothing real. A blended AI-referral figure on a mixed DTC and wholesale business is a DTC figure wearing a company-wide label.
This matters most for the merchants furthest along. The businesses with the largest wholesale share are usually the ones with the most established brands — the exact profile a model is most likely to recommend, and the exact profile whose highest-value revenue the model cannot reach.
To be explicit about the epistemics: this is a structural argument, not a measurement. Every number above belongs to Ethercycle and Shopify and describes storefront traffic. We have no wholesale AI-referral data, and neither, as far as we can find, does anyone who has published on this. We are reasoning from the mechanism to a prediction, and we would genuinely like to see it tested against data segmented by DTC and B2B rather than blended. If someone runs that analysis and the multipliers do carry over, the argument here is wrong and worth knowing about.
Where AI does reach wholesale
None of this means AI is irrelevant to wholesale, now or later. It means the entry point sits somewhere different from where the DTC analyses have been looking.
AI is far more likely to influence the top of a wholesale funnel than to close a wholesale order. A buyer asking a model who supplies a category wholesale in their region is a real, common query. So is a merchant asking a model how to run B2B pricing on Shopify. Both are answered from public, un-gated surfaces: a supplier page, a terms and MOQ page, category and documentation pages, anything a crawler is actually allowed to read.
So the shape is: discovery and qualification can move to the model, and the transaction stays where it already is. The gated portal remains the place where an account is approved, a price is applied and an order is placed, and that part stays human and stays private. The public surface is not the shop. It is the front door, and at the moment most wholesale operations do not have one.
What to actually do about it
Keep something public
If the entire B2B presence lives behind a login, it is invisible to crawlers and to models alike. A public page describing who you supply, typical minimum order quantities, lead times, territories and how to apply for a trade account gives a model something to read without exposing a single negotiated price. This is the cheapest possible fix and most merchants running wholesale on Shopify have not done it, because the portal was treated as the whole of the B2B presence rather than the private half of it.
Structure it
Clear headings, direct answers to the questions buyers actually ask, and structured data where it genuinely applies. Models read pages the way crawlers do, and a page that answers a question in the first sentence of its section is more quotable than one that builds to a conclusion.
Do not chase the 2.5x
It is not available on the wholesale side, and importing a DTC benchmark into a B2B forecast produces targets nobody can hit. Measure your own channels, segmented, and compare them to themselves over time.
Watch the base rate, not the multiple
0.1% growing 6x is 0.6%. That is not a channel to restructure a business around this year, and it is also not a channel to dismiss, because compounding at that rate changes the arithmetic within a small number of years. The correct posture is a cheap public surface now rather than a strategy now.
We build wholesale infrastructure for Shopify merchants at Hasil, which is why the question of what happens to gated revenue in an AI-mediated funnel is one we keep returning to. We do not have the data to settle it. We think the mechanism is clear enough to plan around.
Frequently asked
- How much ecommerce traffic actually comes from AI today?
- In Ethercycle's State of Ecommerce 2026 study — 94 million sessions across ten Shopify stores and $274m in revenue — AI accounted for roughly 0.1% of traffic. That base rate is theirs, not ours, and it is the context missing from most coverage of AI referral multipliers.
- Why do AI referrals convert better than other channels?
- Because the model does the pre-qualification. It interprets the need, chooses a product and delivers the visitor straight to the relevant product page, collapsing discovery and consideration into one off-site step. Ethercycle measured 2.56x the site-wide checkout rate for AI visitors landing on product pages; Shopify cited roughly 2.5x on its Q2 call.
- Will AI referrals help my wholesale revenue?
- Our view is that they will not, in the direct way they help DTC, because the mechanism depends on one decision maker, one session, a visible price and a discovery moment — none of which describe a wholesale purchase. This is a structural argument rather than a measurement; we have no segmented wholesale AI-referral data and would like to see the hypothesis tested.
- Can an AI model see my B2B catalogue?
- Not if it is behind a login, which it should be. Crawlers cannot authenticate into a wholesale portal, so gated products and negotiated prices are unreadable to models. That privacy is the point of the portal; the trade-off is invisibility to AI-mediated discovery.
- What should a wholesale merchant publish publicly?
- A page describing who you supply, typical minimum order quantities, lead times, territories and how to apply for a trade account. It gives models and crawlers something readable without exposing customer-specific pricing, and it is usually the only public B2B surface a merchant is missing.
Want this working on your store?
Hasil ships wholesale portals and Odoo ↔ Shopify sync. 14-day free trial.
Start free trial