Google wants the cart. Shopify welcomes the agent. Who gets the customer?

Google and Shopify are making AI-assisted shopping easier. For retailers, winning the order may not mean becoming the customer's first choice next time.

The aim is not to make comparison harder. It is to make the retailer worth choosing when comparison becomes easier.

Author: Ian Ash

Published: October 15, 2026

Category: Analysis

Retailers could become easier to buy from and harder to remember. Google’s US rollout of Universal Cart, reported on 15 October, and Shopify’s expansion to shopper-selected AI agents make more of the shopping journey possible outside a retailer’s own website. They raise a question that goes beyond AI visibility: after an assistant helps someone buy, what gives that customer a reason to return to the seller rather than start again with the assistant? [1] [2]

For brands investing in answer engine optimisation, appearing in a recommendation remains valuable. Turning that recommendation into a profitable purchase, and a reason to choose the brand again, is a different job. The developments make that strategic tension more concrete, but they do not establish how quickly shoppers will adopt these routes or how loyalty will change.

Two approaches to the same shopping journey

The Verge reports that Universal Cart is rolling out in the US across Google Search and the Gemini app. It saves products from different retailers in one place, sends price-drop and restock notifications, and surfaces promotional codes. The report describes payment through Google Wallet or transfer to a retailer’s own cart. Google’s original May announcement describes Google Pay checkout with participating brands or a handoff to the merchant’s site. A shared place to save products should not be confused with guaranteed checkout for every retailer or a single combined transaction across all sellers. [1] [3]

The chronology matters. Google introduced Universal Cart on 19 May; this week’s reporting concerns its rollout, not a newly invented shopping concept. YouTube and Gmail integration remain planned, with no timeline announced in The Verge’s October report. Retailers outside the reported US rollout should not assume that the same consumer experience is available in their market. [1] [3]

Shopify is taking a different route. The Verge’s 13 October report describes shoppers connecting chosen agents with Shop accounts, purchasing through Shop Pay and receiving orders back in the Shop app. Shop’s own setup page offers personal-shopping connections and documents explicit purchase approval covering the merchant, items, quantities, current total and payment instrument. This is not evidence that consumers routinely give agents unlimited buying authority, or that every agent, merchant and product is compatible. [2] [4]

Google is creating a persistent place to organise purchase intent. Shopify is opening more of its commerce capabilities to agents shoppers choose. Both routes could deliver business to a merchant. Neither, by itself, guarantees that the merchant becomes the customer’s preferred starting point for the next purchase.

Being the seller is not the same as being the starting point

“Who owns the customer?” is useful shorthand, but the relationship is not a single possession. One company can help someone discover a product, another can remember preferences, and another can provide the checkout. The retailer can still earn trust through fulfilment, service and the product itself.

Google explicitly states that the brand remains the merchant of record, regardless of whether checkout happens through Google or on the merchant’s site. The retailer remains the seller; a shopping intermediary becoming more influential does not erase that role. Merchant-of-record status should not, however, be read as proof of exclusive influence over future choices, unrestricted customer data or permission for ongoing marketing. Those questions require the applicable channel terms and customer consent. [3]

A retailer could retain the transaction record while losing the habit of being consulted first. It could also gain a customer who would never have discovered it through conventional browsing. Opportunity and dependency can grow together. The practical question is where that customer returns first, not simply which company processed the last order.

AI could separate loyalty from convenience

Consider a hypothetical shopper who repeatedly buys household products from one retailer. Some repeat purchases may reflect genuine preference: dependable delivery, helpful service or confidence in the product. Others may reflect a familiar account, saved shopping list and easy checkout. Those reasons can coexist, but they are not interchangeable.

An assistant that makes comparison and ordering easier across sellers could weaken convenience-based retention. A request such as “Find the best delivered price for the exact product I usually buy” could reopen a decision that previously went unchallenged. This is an editorial hypothesis about reduced decision effort, not a measured loyalty effect from either platform’s announcement.

The reverse is also plausible. An assistant instructed to buy a particular brand could make it easier to remain loyal to that brand, while changing the retailer that supplies it. A retailer with distinctive service, reliable fulfilment or genuinely valuable membership benefits could remain the best choice after comparison. The distinction is between “I want this brand” and “I usually buy here”. Retailers need to understand how much of their repeat business depends on each.

This extends the question explored in AEO Updates’ analysis of AI and the future of customer loyalty: what happens when software makes reviewing an existing choice easier? In retail, that review may concern the seller of an already-preferred product, rather than whether the customer wants a different brand at all.

The offer an agent sees may differ from the storefront

A smaller Shopify update illustrates a practical information problem. On 9 October, Shopify announced that compare-at prices are now shared with AI shopping agents through Shopify Catalog. Previously, agents reading Catalog saw current prices without the markdown context displayed in the storefront. Merchants can control whether the comparison price is shared through a setting. [5]

That is a specific example of an offer being represented differently to a human shopper and an agent. It does not prove that every assistant displays markdowns, uses them to rank options or treats a comparison price as evidence of value. A reference price also does not establish that every customer previously paid it, or that the resulting offer is the most competitive.

Retailers should test other consequential parts of their own offer: product variants, delivery estimates, total delivered cost, returns and membership conditions. A persuasive product page and a complete commerce feed serve different purposes. Both should represent the offer accurately, including the limitations that matter to a buying decision.

Test the purchase, then investigate the next visit

The immediate response should be a focused review involving marketing, ecommerce, customer service and analytics. Choose a handful of real customer needs and follow each from an AI recommendation through to an authorised order. Record missing information, mistaken product matches, unclear conditions and checkout failures. As the retailer guide to selling through AI shopping agents explains, discovery, checkout and a completed order are separate stages. A brand mention alone does not establish that the buying journey works.

Then ask repeat customers why they return. Do they want a particular product, trust the seller’s service, value a membership benefit or simply find reordering easy? Include cases where the current retailer is genuinely the best option, rather than designing research to assume that AI must recommend switching. Those answers help distinguish an information problem from a weak offer, and identify which advantages deserve investment.

Where attribution allows, compare conversion, contribution after channel costs, returns and repeat purchasing across AI-assisted and other channels. Keep incomplete attribution visible. Differences between channel customers may reflect who uses the channel rather than an effect of AI; controlled or longitudinal evaluation is needed before claiming incremental growth or a causal retention effect. Visibility is evidence of discovery, not proof of profitable loyalty.

Give the customer a reason to remember the seller

Accurate fulfilment, useful advice, easy problem resolution and distinctive products can provide reasons to choose a seller again after an intermediary introduces the customer. Retailers should investigate whether these benefits are understood by customers and accurately represented to assistants. Seek appropriate permission for further communication rather than assuming that an intermediary-assisted purchase grants unrestricted access to the customer.

The aim is not to make comparison harder. It is to make the retailer worth choosing when comparison becomes easier. A useful test of progress is whether customers still name a defensible reason to prefer the seller after the convenience of searching and ordering has become less distinctive.

AEO Updates Takeaway

Google and Shopify are expanding the routes through which a shopping journey can happen beyond a retailer’s website. That does not establish that brand loyalty will disappear, or that merchants inevitably lose their customer relationships. It makes the difference between winning an order and earning the next visit harder to ignore. Retailers should welcome qualified demand, verify that their offer survives the agent-mediated journey accurately, and investigate whether the purchase gives the customer a stronger reason to choose the seller again or simply a stronger reason to return to the assistant.