Amazon Blocks Shopping Agents, Shopify Opens Up: Only One Strategy Works for Most Retailers
Amazon has shut out Meta's shopping agent while Shopify wires its merchants into ChatGPT, Copilot and Google. Blocking suits a retailer whose shoppers start in its own app, and the retailers copying Amazon mostly do not have that.
Amazon has shut out Meta's new Muse shopping agent. That is the plainest statement yet of what a large retailer thinks an AI intermediary is: a stranger standing between the store and its customer. Shopify, whose merchants sell against Amazon, has reached the opposite conclusion and is wiring its stores straight into the agents Amazon turns away.
The Wall Street Journal's CMO Today column laid out the split on Friday. Amazon aims to block shopping agents entirely. Kohl's lets them look but prevents agentic purchases. Tapestry, owner of Kate Spade and Coach, lets agents browse and stops there, though it is optimising its systems for agentic shopping so that it is ready if it opens that lane. Gap, Walmart and Best Buy have publicised partnerships with Muse, and QVC and HSN broadly welcome agents to complete purchases and send their catalogues to OpenAI and Google.
The sorting variable is who owns the customer, and it decides which policy is rational. Amazon can say no because shoppers who want Amazon open Amazon; an agent that cannot get in loses the user, not the other way round. A retailer whose shoppers start in a search box or a chat window is in a different position. It depends on being found, and a block that costs Amazon nothing costs a department-store chain its place in the answer. Hesitant retailers cite data control and customer experience, which are fair worries and the same reasons Amazon gives. Only Amazon can afford to act on them.
Shopify sits at the far end of the open side. It says its Agentic Storefronts give merchants out-of-the-box access to ChatGPT, Microsoft Copilot and AI Mode in Google Search, announced on March 24. It co-developed an open Universal Commerce Protocol with Google, announced January 11, and said on May 11 that AI-referred shoppers convert better. The logic suits a platform whose merchants mostly lack Amazon's audience. Shopify earns when they sell, and every extra channel that sends them a shopper helps.
The obvious objection is volume. The WSJ reports there is no evidence yet that significant numbers of people are sending agents out to buy, and Tapestry's Yang Lu says customers are adopting agentic commerce much more slowly than other AI tools. Research from Billions Network suggests most large retailers' pages probably block agents by default anyway, and QVC says its old bot protections may be turning agents away by accident, so some of the blocking is plumbing and not strategy. All of that argues for low expectations about near-term sales. It does not argue for blocking. With volumes near zero, being open costs almost nothing, and being absent when volumes arrive costs a retailer the position it would have built.
Canvas is the other half of Shopify's bet. Shopify launched it on Thursday, according to its changelog, as a design surface for merchants: every page of a store sits in one interactive infinite workspace, and a merchant can click an element to edit it or tell Sidekick, Shopify's AI assistant, what to change and watch the result appear in real time. If agents bring shoppers, the storefront they land on is worth designing, and Canvas lowers the cost of doing it well. It is a tool for human store owners, and its limits show at launch: it does not yet support third-party Theme Store themes, Markets, rollouts, translations, or app blocks and embeds, and a theme edited in Canvas stops receiving theme updates. Canvas reaches existing stores over the coming days, so the first Sidekick-built storefronts should be visible within weeks.
The market is paying for the optimism. Shopify traded at $151.04 in Friday morning trading, up 1.31%, for a market value of $194.3 billion and a price to earnings ratio of 101.8, inside a 52-week range of $94.00 to $182.19. At that multiple, investors are paying for the idea that Shopify's merchants are the ones agents will want to shop with, which rests on Shopify's own finding that AI-referred shoppers convert better and on nobody having yet measured agent behaviour at scale.
Shein shows what the market pays when the customer relationship turns out to be borrowed. Goldman Sachs bought about 42 million Shein shares last month, roughly 13% of the initial float and about $220 million, according to a Hong Kong Stock Exchange filing cited by the Financial Times. It paid between HK$35.90 and the HK$48.56 listing price, an average of HK$42.23, under its stabilisation role in the offering, a routine part of an underwriting that shows a trading profit of about $34 million at the average price. Under the standard agreement, stabilisation profits are shared among the underwriting banks, although in some deals, such as Deliveroo's disappointing London listing in 2021, the banks agreed in advance to hand them to the company. The long-awaited $26 billion IPO was labelled a "disaster" after shares fell as much as 10% on day one, and the stock is down 38%.
Shein's demand came, in the FT's account, from pandemic lockdowns and social-media hype around "Shein hauls", which earned it a valuation of about $100 billion in a 2022 private round. The market now values it at $16.3 billion against $15 billion of cash at June 30, which leaves roughly $1.3 billion for the operating business. A retailer that never owned the route its customers took to reach it is worth about its cash. The retailers blocking agents are betting that they own theirs, and Amazon is right. Most of the others should check whether they do.