The next visitor deciding whether to recommend your product may not be a person at all, and it will judge your pricing by whether it can understand it.
Something worth paying attention to started this spring: AI shopping assistants began recommending real products to real shoppers. The consumer versions, including shopping features inside ChatGPT, are early and imperfect, and we want to be careful not to oversell what they can do today. But the direction is clear enough to think about now, and the practical question for a store owner is simple. If an assistant is reading your listing to decide whether to suggest it, can it understand your pricing?
For a lot of stores, the honest answer is not really. And that is the part worth fixing, because the fix is good practice regardless of how big this trend gets.
What is actually happening
An AI shopping assistant sits between the shopper and the store. A person describes what they want in plain language, the assistant researches options, and it points them toward specific products. To do that, it reads whatever it can: the product title, the description, the price, the reviews, the structured data behind the page.
That last point is the one merchants underrate. A human shopper will tolerate a messy product page, hunt for the real price, and mentally apply the "buy 3 save 10%" banner. An assistant will not work that hard. It represents what it can read cleanly, and it tends to favor listings it can describe with confidence. Ambiguity is a tax, and the assistant passes that tax on to you by recommending someone clearer instead.
The pricing patterns that confuse a machine
A few common setups are fine for people and hard for assistants.
Deals that only exist at checkout. If the real price only appears after a code is applied or the cart hits a threshold, the assistant reading your product page sees the higher number. The value you are offering is invisible at exactly the moment a recommendation gets made.
Discounts with no numbers. A "bulk savings available" badge with no per-unit prices gives an assistant nothing concrete. It cannot cite a value it cannot read.
Inconsistent pricing across variants. If the same product shows contradictory prices depending on where you look, an assistant has no reliable figure to represent, so it tends to skip the item rather than guess.
None of these are new problems. AI just makes them more expensive, because a confused human sometimes buys anyway and a confused assistant simply moves on.
The measured response
We are not going to tell you to rebuild your store around a spring-2025 trend. It is too early for that, and anyone claiming certainty about where agentic commerce lands is guessing. But there is a version of getting ready that costs you nothing if the trend stalls and helps if it grows, because it is just clearer merchandising.
State price per unit. A volume discount shown as a concrete unit price at each quantity ("$20 each, $17.60 each at 6") is legible to a shopper and to a machine. Both can reason about it. A vague banner cannot be reasoned about at all.
Prefer automatic, visible discounts over code-gated ones where it makes sense, so the price a reader sees is the price the customer pays. There are good reasons to use codes, but understand that a code hides value from anything reading the page cold.
Keep margin discipline. This is the quiet one. If assistants start sending you traffic through channels you do not control, you want every discount you run to be defensible on margin, not a promotion you were relying on nobody noticing at scale. Watching margin in profit analytics matters more, not less, when the source of a sale is a recommendation you did not place.
A small example
Two stores sell the same $20 product with a buy-6 deal.
Store A shows a banner: "Bulk discounts available in cart." The product page price is $20. An assistant reads $20, sees no concrete deal, and has nothing to highlight.
Store B shows a small tier table: 1 unit $20, 3 units $19 each, 6 units $17.60 each. An assistant reads real per-unit prices at real quantities and can tell a shopper exactly what buying more saves.
Same underlying offer. One is legible and one is not. That is the entire difference, and it is a merchandising decision, not an AI project.
Where this goes next
This is the first time we have written about AI shopping on this blog, and we are keeping it grounded on purpose. What we can say now is narrow: pricing that is clear to a person is becoming clear to a machine too, and the stores that already merchandise their volume discounts transparently are, almost by accident, the ones best positioned for whatever this becomes.
Shopify's summer Edition lands in a couple of weeks and is expected to lean hard into AI, so we will have more to say once we can react to something concrete rather than a trend line. When we do, it will be in our Summer '25 Editions breakdown. For now, the homework is unglamorous and useful: make your pricing something a reader can actually understand, and make sure the discounts behind it still hold their margin.
