The Maze: AI shopping assistants are becoming the new shelf for product research. They can rank products, show cards, compare options, and push users toward a decision. But the transaction still sits mostly with retailers. The public EMARKETER report frames the split cleanly: shoppers may ask ChatGPT, Google, or Copilot what to buy, but price, stock, loyalty, checkout, delivery, and returns still depend on merchant infrastructure.
The three big assistants now look more similar than different at the research layer. ChatGPT, Google, and Microsoft Copilot all return product recommendations with enough commerce context to feel useful: prices, sellers, ratings, product cards, comparison logic, or recommendation lists. That matters because discovery is no longer just a search-results page. It is becoming a conversational shortlist. The user sees fewer options, so the quality of product data, feed structure, and brand visibility inside assistant answers becomes a real digital shelf problem.
Checkout is the fault line. In EMARKETER's feature comparison, ChatGPT has no in-chat checkout, while Google and Copilot show checkout capability. That sounds like a small product-detail difference. It is not. Checkout pulls in identity, payment, loyalty, tax, delivery promise, returns policy, fraud controls, and customer service. AI can summarize a product fast. Retailers still carry the operational debt of actually selling it.
Loyalty and inventory turn assistant commerce from clever demo to retail system. ChatGPT can link loyalty or membership only within third-party apps. Google has broader loyalty linking. Copilot is limited to select retailers. Fabio Plebani's post adds the commercial caveat: OpenAI has acknowledged pricing and inventory accuracy of only 64%. If one in three availability or price signals can be stale, the assistant is useful for inspiration but risky for a buying decision. In commerce, wrong confidence is worse than slow search.
Ads are already part of the story, even before the transaction fully moves. ChatGPT and Copilot both show ads in the comparison, while Google is partial across AI Mode and AI Overview. That creates an awkward middle phase. Platforms can monetize shopping attention before they own the full purchase. Retailers may still close the sale, but assistants can tax the upper funnel, shape the shortlist, and decide which brands are even considered.
The retailer advantage is not a prettier interface. It is trusted data. The strongest comment thread under the post focused on data quality, local inventory, regional pricing, shipping data, and structured feeds. That is the unglamorous base layer. Retailers that keep product data current can appear in AI answers with confidence. Retailers with messy feeds may disappear before the customer reaches their site. AI does not remove ecommerce plumbing. It makes bad plumbing visible.
Why it matters: Agentic commerce will not arrive as one clean platform takeover. It will arrive as a stack. AI assistants will influence discovery, comparison, and persuasion. Retailers will keep controlling fulfillment, loyalty, payment, and trust. The margin battle moves upstream first. Whoever owns the product data, not just the chatbot window, gets leverage over the next shopping journey.
Sources: EMARKETER report | EMARKETER feature comparison | LinkedIn post
