When the Cart Became the Control Layer


OZ Signals

May 26, 2026

When the Cart Became the Control Layer

Issue 9 tracked the rise of the implementation layer, where agencies, enterprise systems, storefront builders, and deployment partners started making AI commerce installable inside real businesses. That mattered because AI commerce was no longer only a platform announcement. It was becoming something merchants could actually connect, configure, and operate.

Issue 10 moves one layer deeper. The week of 19 May to 25 May showed that the next fight is not just implementation. It is control at the decision surface. Google’s I/O and Marketing Live updates revealed a commerce environment where the cart, product feed, ad unit, checkout protocol, payment mandate, and merchant dashboard are being pulled into one machine-readable loop.

The important shift is this: the cart is no longer the end of the shopping journey. It is becoming the place where comparison, compatibility, price logic, loyalty, payment rules, merchant eligibility, and purchase authority come together before the human has fully decided. That changes what merchants must control. It also changes where commerce power sits.

Universal Cart turns the basket into a reasoning system

Google introduced Universal Cart on 19 May 2026 as an intelligent shopping cart that works across merchants and Google surfaces including Search, Gemini, YouTube, and Gmail. It can monitor price drops, track stock, surface price history, flag product incompatibilities, suggest alternatives, and use Google Wallet data such as payment perks, loyalty information, and merchant offers before checkout. Google says the cart will roll out across Search and Gemini in the U.S. this summer, with YouTube and Gmail to follow.

Structurally, this moves the cart from a passive holding area to an active decision layer. The old cart stored intent after the shopper selected. The new cart shapes intent while the shopper is still deciding. This matters because compatibility, loyalty, offer logic, payment method advantage, delivery fit, and price movement all become machine-evaluated inputs inside the purchase path.

What changes is the merchant’s relationship with the basket. A product no longer competes only on a product page or search result. It competes inside a live reasoning environment that can decide whether the item fits the rest of the basket, whether another merchant offers a better configuration, and whether a payment method makes one option more attractive than another.

The second-order implication is uncomfortable for brands. Merchandising loses some of its final-mile power when the cart can correct, compare, and redirect the shopper. The product that wins may not be the best-branded product. It may be the product whose data, price logic, payment perks, stock status, and compatibility signals are easiest for the cart to defend.

Break: The cart is no longer where demand is captured. It is where demand gets edited.

Source: Universal Cart

Conversational Attributes make product data a sales interface

Google surfaced Conversational Attributes for Merchant Center, describing them as data attributes designed for discovery in conversational commerce experiences such as AI Mode. Google says these attributes help answer more detailed shopper questions, complement existing product feeds, support recommended complementary products, and let merchants reuse product data already prepared for other platforms.

This is not a small feed update. It is a change in what product data is expected to do. Traditional product feeds were built for listing, matching, and ranking. Conversational attributes are built for explanation. They help AI systems answer questions like whether a product fits a use case, pairs with another item, solves a specific constraint, or belongs in a recommended bundle.

What changes is the commercial value of hidden product knowledge. Details that once lived in copy, customer service scripts, PIM fields, Amazon listings, training decks, or retailer-specific feeds now become inputs for agentic recommendation. A missing attribute is not just an operational gap. It can become a lost sale because the system cannot confidently explain why the product belongs in the answer.

The second-order implication is that catalog governance becomes a growth function. Merchants will need to treat attributes like pricing, inventory, and performance data. The businesses that win will not simply have better products. They will have products that can be read, reasoned over, compared, and defended inside conversational decision systems.

Break: Product descriptions stop being persuasion when machines need structured proof.

Source: Attributes

UCP expands from checkout plumbing into category infrastructure

On 20 May 2026, Google said it is expanding Universal Commerce Protocol features across Google, including UCP-powered checkout through Google Pay, YouTube Shopping ads, Direct Offers, and new categories such as hotel booking and local food delivery. Google also said UCP-powered checkout will roll out in Canada and Australia, followed later by the U.K.

This matters because UCP is no longer only about making checkout smoother. It is becoming a common transaction language across surfaces, countries, ad formats, and verticals. Once a protocol moves from retail checkout into travel and food delivery, it stops looking like a feature and starts looking like a commerce rail.

What changes is the boundary between discovery and transaction. Search, Gemini, Maps, YouTube, ads, checkout, hotel booking, and food delivery begin to share a common transaction pathway. That means the old merchant model of separate journeys, separate surfaces, and separate funnel logic becomes harder to defend. The system wants one readable path from intent to execution.

The second-order implication is that protocol readiness becomes market access. Merchants and platforms that can speak the transaction language will enter AI-mediated demand flows faster. Those that cannot may still appear online, but they will be harder for agents to complete, book, bundle, or recommend.

Break: Being discoverable is not enough when the system also needs you to be executable.

Source: UCP Expansion

AI ads begin carrying their own explanation layer

Google announced new AI ad formats for Search and AI Mode on 20 May 2026, including Conversational Discovery ads, Highlighted Answers, AI-powered Shopping ads, and Business Agent for Leads. Google says Gemini will synthesize product information, generate tailored explanations, and place independent AI explainers beside advertiser creative while keeping ads labeled as sponsored.

This changes the role of advertising inside commerce. The ad is no longer only a placement, message, or creative asset. It becomes a machine-generated answer that interprets the product for the shopper’s exact context. That is a structural shift because persuasion starts moving from brand-crafted claims to AI-mediated explanations built from available product, website, feed, and campaign data.

What changes is the creative bottleneck. Brands used to optimize copy for attention. Now they must optimize evidence for explanation. If the system is generating the reason a product is relevant, the merchant’s job is to make sure the system has enough accurate material to build a useful and trustworthy answer.

The second-order implication is that advertising and product data begin to merge. Weak data will produce weak explanations. Inconsistent claims will create fragile recommendations. The strongest ad account may not be the one with the best creative team. It may be the one with the cleanest product truth.

Break: Paid visibility loses power when the explanation layer cannot justify the recommendation.

Source: AI Ads

Ask Advisor turns merchant operations into an agent-managed loop

Google introduced Ask Advisor on 20 May 2026 as a cross-product AI agent that connects Google Ads, Google Analytics, Google Marketing Platform, and soon Merchant Center. It can pull product details from Merchant Center, help set up campaigns in Google Ads, surface insights from Ads and Analytics, understand business goals, explain what worked, and recommend next steps. It is currently in beta for English-language accounts, with more features rolling out later.

This is structurally important because the merchant console is becoming agentic. The operator is no longer only moving between dashboards, reports, feeds, and campaign tools. The system is beginning to coordinate those functions through one agent that understands goals, data, listings, performance, and campaign execution.

What changes is the operating model for small and mid-sized merchants. A founder or marketer can ask for growth, and the agent can translate that into campaign setup, product-feed use, performance diagnosis, and next action. That compresses the distance between strategy and execution. It also gives the platform more influence over how merchants define problems, choose actions, and allocate spend.

The second-order implication is that merchant capability becomes partially platform-shaped. Businesses that rely on these agents may execute faster, but they may also start thinking in the platform’s language. The dashboard becomes less like a tool and more like an operating partner.

Break: Merchant autonomy changes when the platform starts diagnosing the business and executing the fix.

Source: Ask Advisor

The System That Is Emerging

The hidden system this week is not “AI shopping.” That framing is too small. What is emerging is a commerce control layer where carts reason, feeds explain, protocols execute, ads justify, payments authorize, and dashboards act. The important movement is that Google is not only improving individual shopping features. It is connecting multiple commercial functions into one machine-readable operating path.

Control is moving away from the merchant website as the main place where persuasion, comparison, checkout, and measurement happen. It is moving toward the surface where intent is first expressed and where the system can keep reading that intent across search, chat, video, maps, ads, wallet, and merchant tools. The merchant remains the seller of record in many of these flows, but the system increasingly controls the conditions under which the sale becomes visible, explainable, executable, and measurable.

For operators, this means the old commerce stack needs a new readiness layer:

  • Catalogs must explain, not just list.
  • Carts must be treated as decision engines, not storage.
  • Ads must carry evidence, not just messaging.
  • Checkout must be protocol-ready, not just payment-ready.
  • Merchant tools must be governed, because agentic recommendations can shape business decisions upstream.

Core Truth: The merchant that cannot make its products explainable, executable, and governable inside machine-led surfaces will lose control before it loses the customer.

This is why the week matters. The shift is not that Google added a smarter cart or better ads. The shift is that the commercial system is learning to complete more of the shopping journey without returning power to the merchant’s owned environment. The website still matters, but it is no longer the only place where the market decides.

Tool of the Week Google Merchant Center Conversational Attributes

The most structurally relevant tool this week is Conversational Attributes in Google Merchant Center. It matters because it gives merchants a practical path to prepare product data for AI Mode and agentic discovery. Instead of treating product information as static catalog content, merchants can begin shaping the fields that AI systems use to answer nuanced questions, recommend complementary products, and explain why a product fits a shopper’s intent.

At a system level, this tool sits directly inside the new control layer. It turns product truth into machine-readable commercial evidence. For merchants, the first job is not to “optimize for AI” in a vague way. It is to audit which product attributes help a system answer real shopper questions, then make those attributes complete, consistent, and current across Merchant Center and other selling surfaces.

Source: Attributes

Trend to Watch The cart becomes the new marketplace interface

The early pattern to watch is the shift from marketplace pages to marketplace carts. A marketplace used to be a place where shoppers browsed options. In this new model, the cart itself can compare merchants, combine products, check compatibility, apply loyalty logic, surface payment advantages, and complete or redirect checkout. That makes the cart a new competitive arena, especially for categories where products are bought in combinations, like electronics, beauty routines, home projects, travel, food delivery, and replenishment.

This trend will matter because merchants may soon need to optimize not only for search ranking, ad performance, and product page conversion, but for basket compatibility. A product may win because it fits another product, qualifies for a promotion, works with a payment method, ships within a constraint, or gives the agent a cleaner reason to recommend it. That is a different kind of commerce competition.

Issue 10 shows that AI commerce is moving from deployment into control. The previous layer made AI commerce installable. This layer makes commerce itself increasingly machine-managed at the point where intent, product data, offers, ads, checkout, payments, and operations meet. The merchant is still present, but the decision environment is being rebuilt around systems that can read more context than a website, act across more surfaces than a store, and shape more of the path before the buyer reaches the brand.

OZ Signals will continue tracking the next layer because this is where the structural advantage starts to separate. The obvious conversation will be about better shopping experiences. The more important question is who controls the rules those experiences follow. The next phase of AI commerce will not reward businesses that only look good to humans. It will reward businesses that can be understood, trusted, executed, and defended by machines.

Box Hill (Sydney), NSW 2765, Australia
Unsubscribe

OZ Signals

OZ Signals is a weekly intelligence briefing on how AI is restructuring commerce systems. Built for founders, operators, and decision-makers who want high-signal insights, not noise.

Read more from OZ Signals
Minimal editorial image showing AI commerce moving from concept into implementation through storefronts, agencies, assistants, and operational systems

OZ Signals May 19, 2026 View in browser AI Commerce Is Moving Into the Implementation Layer Last week’s issue showed that AI commerce is moving beyond checkout into fulfillment, logistics, service, and post-purchase systems. This week explains the next layer: implementation. The market is no longer only experimenting with AI commerce. It is beginning to build the infrastructure required to deploy it across real businesses. That shift matters because most retailers are not limited by awareness...

Minimal editorial image showing an AI commerce loop moving from intent to payment, then into fulfillment, service, clienteling, and post-purchase support.

OZ Signals May 12, 2026 View in browser The Transaction Is No Longer the End of Commerce Last week’s issue showed that the payment layer is becoming the execution layer. Once AI systems can select what to buy, the next structural question is how they receive controlled authority to spend. That layer matters because agentic commerce cannot scale if agents can recommend products but cannot safely complete the transaction. This week moves one step further. Once payment becomes executable, the...

Minimal editorial image showing an AI agent receiving controlled payment authority through a wallet, with rails branching into cards, mobile wallets, procurement, and stablecoin settlement.

OZ Signals May 05, 2026 View in browser The Payment Layer Is Becoming the Execution Layer Last week’s issue showed that selection is moving from market to model. Demand is no longer entering commerce only as traffic. It is entering as structured intent, filtered by systems before merchants get to compete. That explained who gets seen. This week moves one layer deeper. Once the system has interpreted demand and selected an option, the next question is not discovery. It is whether the agent can...