Amazon vs Shopping Agents: What a Small Store Does

Amazon has spent the autumn building walls while Shopify has been taking them down. That contrast is the most useful thing a small merchant can learn from the shopping agents story, because it clarifies which side of the fight your store actually sits on.

The short answer: Amazon is trying to keep agents out and route shopping through Alexa, while Shopify opened catalogue and checkout to any agent. Neither side is fighting over your store, and neither has to. What matters is that both are fighting over the moment of discovery, and a small store wins that fight with structured product data and nothing else.

What Amazon actually did

The sequence, reported by the Wall Street Journal on 1 November and summarised by PYMNTS:

When we tested product pages against agent queries this autumn, the pages that scored best were the ones that named a limitation, which is why the data advice below leans that way.

Paul Graham's reaction, posted on X on 7 October, according to reporting by PYMNTS, is the line the story keeps returning to: "Amazon banning agents is the first opportunity I've seen since Amazon was founded for a startup to create an Amazon competitor. People will want agents to buy stuff for them."

The number that undercuts the confidence

Jassy says agents are not a real shopping behaviour yet. The research partly agrees and partly does not.

According to PYMNTS Intelligence's study of the 2026 shopping season, Amazon captures 59% of the most recent AI-assisted purchases, and accounts for 29% of research starts but 49% of main purchase completion.

Read that carefully. AI-assisted shoppers still end up buying on Amazon roughly half the time. The agents are researching more broadly than they are transacting, and the transaction is still flowing to the retailer with the fastest delivery and the simplest returns.

So Amazon's wall is not defensive paranoia. It is protecting a position that is currently winning.

Why that is not the whole story

Here is the part that matters for a store selling something Amazon does not stock, or stocks badly.

If 59% of AI-assisted purchases finish on Amazon, then 41% finish somewhere else. That is a large minority, and it is the minority a Shopify merchant competes for. The question is not whether you can beat Amazon at agentic commerce. It is whether you are legible enough to be the alternative the agent offers when Amazon is not the right answer.

That legibility is a data problem, not a platform problem. It is also, per the same body of research, where almost everyone is failing: only 15% of merchants hold structured product data agents can use, as we broke down in the agent readiness data.

The three-way split you are actually in

Note the asymmetry. Amazon can choose to be closed because it has demand. A small store cannot afford to be closed, because it does not have demand to protect. Opening up is not a strategy for you so much as the only option available.

That is also why the panic about Amazon blocking agents is misplaced. If Amazon will not let an agent check out, the agent has to send the shopper somewhere, and "somewhere" is an open catalogue. A store that is ready for that is competing for the 41%, not against Amazon.

What to do this week, not this quarter

1. Finish your product attributes. Material, measurements, compatibility, contents, care. This is the same work that gets you into an editorial collection and the same work that gets you recommended by an agent.
2. Say who the product is not for. Agents answer scoped questions. A page that admits "not suitable for X" is easier to match than one that claims universal fit.
3. Check you are reachable without a key. Shopify's agent infrastructure is already open on every store, including Hydrogen and Liquid storefronts alike. Shopify's Shop.dev connection needs no API key, which means the barrier is your catalogue rather than your integration, and our storefront comparison covers why that is true on both Hydrogen and Liquid.
4. Do not chase the Amazon fight. Nothing you do changes what Amazon does. Everything you do changes whether an agent can describe your product accurately.

Where our view is limited

We sell storefronts, so "make your product data better" is a conclusion that happens to serve us. Discount it accordingly.

There is also a real argument on Amazon's side that we cannot dismiss. If agents strip out the browsing and the ads, the storefront experience stops being a marketing surface and becomes a fulfilment interface. Amazon is already excellent at the second thing, and a small store betting on experience may be betting on a surface that agents make less valuable.

And the honest uncertainty: the readiness data says agents are barely transacting. According to Checkout.com, agent-involved transactions sit at 3%. This may all be infrastructure built ahead of demand that arrives slowly, or never.

7-day free trial · all themes included · cancel anytime