If you have ever sold a product that other stores also sell, there is a question worth asking about agentic commerce that almost no one is asking yet: when an assistant finds that product, whose offer does it show?
I went looking for the answer in Shopify's Global Catalog, the cross-merchant index that ChatGPT, Gemini, and Shop search over. What I found is a quiet but important mechanic that behaves a lot like the Amazon buy box, and it is going to matter more every quarter.
Products cluster, offers compete
When the catalog returns results, it does not list every store's listing as a separate product. It clusters them by a Universal Product ID, a single identity for "this product," and then attaches the offers from each merchant that sells it. The assistant sees one product and a set of competing offers underneath it. When it recommends a place to buy, it is choosing among those offers.
I tested this across a range of searches. For commodity and widely-resold goods the clustering is very real: a search for a multipack of batteries returned a single product with three different sellers competing on it, and common branded items like insulated bottles and popular sneakers regularly clustered two sellers on one product. For unique, own-brand products it was the opposite. A coffee roaster's house cold brew and a direct-to-consumer brand's flagship item stood alone, one seller, no competition on that identity.
That split is the whole strategy in one observation.
Two games, depending on what you sell
If you resell products that other stores also carry, you are in a buy-box game whether you realized it or not. The assistant has multiple offers for the same identity and will lean on price, shipping, availability, and ratings to pick one. You are not competing to exist in the result. You are competing to be the chosen offer inside it. I even saw cases where a brand's own product was outranked on its own name by a reseller, because the reseller's offer looked better on the signals the catalog weighs.
If you sell unique own-brand products, there is no shared identity to fight over, so the game is different: presence and rank. You are not trying to beat another seller on the same item. You are trying to appear at all for the category searches that matter, and to rank above genuinely different alternatives. The branded-versus-category problem I wrote about recently is the real risk here, not the buy box.
Most catalogs are a mix of both, which means you are playing both games at once and should know which products fall into which.
What decides the offer
For the buy-box game, the levers are the unglamorous ones. Price is the loudest signal, but it is not the only one. Shipping coverage matters, because an offer that cannot ship to the buyer is not a real option. Availability matters, because an out-of-stock offer is no offer. Ratings matter, because they are the trust signal an assistant leans on when it has to choose for someone. And condition matters, because "new" and "secondhand" are filtered separately.
The uncomfortable part is that you cannot see any of this from inside your own store. You can look at your prices all day and never know you are the fourth-cheapest offer on a product an assistant is recommending to thousands of shoppers. The competitive picture only exists in the catalog, and the catalog does not send you a report.
What to do about it
If you resell, audit the products where you share an identity with other sellers and look at where your offer actually lands on price, shipping, and ratings. The fixes are concrete: tighten shipping coverage, keep stock honest, earn and surface reviews, and price with the knowledge that a machine is comparing you instantly.
If you are an own-brand store, do not over-index on the buy box. Spend your effort on category discovery and rank, because that is where your products live.
Either way, keep your catalog data clean and current, because every one of these signals is read from your structured product data, and a machine reads it literally and often.
Where we land
Agentic commerce is not just a new place to be found. For a growing share of products it is a new competition to be chosen, decided on signals you cannot see from your own admin. Knowing which of your products share an identity with other sellers, and where your offer stands on each, is becoming a real merchandising discipline.
That visibility is exactly what we are building toward with AgentReady: not just making your catalog legible to assistants, but showing you where you actually stand inside the Global Catalog, on the products and searches that matter to your store. The first step is being legible. The next is knowing whether you are winning.
Comments
Every comment here comes from a verified email. Write yours, confirm from your inbox, and it's live.
Loading comments…