In the past, shopping searches would direct users to product pages, and subsequent tasks such as price comparison, logging in, filling in addresses, and making payments were still completed by the users themselves. On November 4th, Google announced the integration of direct checkout based on common business protocols into its search AI and Gemini services in Canada. The first batch of partners included lululemon, some Shopify merchants, Shoppers Drug Mart, and Simons. After researching products through these interfaces, users could complete purchases directly from eligible retailers. Although it may seem like just a few fewer page transitions, this actually represents a realignment of control over retail entry points, order assignments, and payment processes.
From "Recommended Products" to "Pending Tasks"
The traditional business logic of search is to display links, which are then followed up by retailers' websites to convert visitors into customers. Intelligent shopping takes this process a step further: the system not only understands "I want a coat suitable for winter commuting in Vancouver," but it can also filter options, compare choices, and complete the checkout process with the user's permission. According to Google, 59% of Canadian holiday shoppers plan to use AI tools to assist with their shopping this season, indicating that conversational interfaces are no longer just an experiment for a minority of people.
The real change is that intentions have been compressed into actions. In the past, users had to open many tabs to repeatedly check sizes, inventory, prices, and shipping details; now, platforms aim to organize these steps within a single continuous conversation. With each reduction in redirects, the likelihood of a transaction occurring increases, but the platforms also gain more data about user preferences, the comparison process, and the timing of purchases. Whoever controls this conversation is one step closer to the moment the transaction takes place.
The general protocol aims to solve the issue of duplicate connections.
The biggest engineering challenge for agents to work across different stores is not the language models, but rather the fact that each store has its own unique products, inventory, payment, and after-sales interfaces. The so-called universal business protocol mentioned by Google aims to enable consumer interfaces, merchant systems, and payment services to use a common language. This protocol covers the discovery, purchase, and after-sales processes, reducing the cost for retailers to develop separate connections for each agent.
This is also an important reason why Shopify merchants have made it onto the first batch of the list. For small and medium-sized businesses, if the integration cost is low enough, they have a chance to be noticed through the new AI entry point, without having to build their own complete dialogue shopping systems. Google plans to provide a simplified integration method from next year onwards, which means that this launch is not a closed experiment, but rather a preparation for serving a larger number of merchants.
Whoever is the seller shall bear the responsibility.
Directly settling the account on the Google interface does not equate to becoming a product seller on Google. According to the official instructions, retailers are still recorded as sellers and can adjust the integration method according to their own needs; payments use the existing payment and delivery information from Google Pay and Google Wallet. This arrangement reserves the experience entry point for Google, while retaining the responsibility for products, order fulfillment, returns, and exchanges with most consumers to the merchants.
This boundary is crucial. If the chatbot recommends the wrong size, ignores delivery restrictions, or misinterprets promotional conditions, users will likely attribute the problem to the dialogue interface, while merchants may end up bearing the subsequent handling costs. Retailers need to ensure that the information displayed by the chatbot is consistent with real-time inventory and terms, and they also need to be aware of which decisions in the order are explicitly confirmed by the user and which are handled automatically by the system. Without clear authorization records, convenience can easily lead to disputes.
Retailers achieve conversion, and also hand over a portion of the front desk operations.
For brands, new entry points bring a dual nature. On one hand, there is a shorter purchasing path and a lower abandonment rate: users can complete transactions right where they have the need. On the other hand, brand websites, display methods, and membership operations may be streamlined. What consumers remember is "buying at Gemini," rather than the carefully designed experience of a particular store. Google indicates that features such as loyalty rewards will be added in the future, which precisely shows that membership relationships remain a core asset that merchants are reluctant to give up.
When retailers integrate, they should not only focus on new orders but also observe average order value, return rates, member recognition rates, and channel growth. If the AI entry point merely shifts users who would have made purchases on the official website to a new intermediary layer, then what merchants gain is not growth, but rather a greater dependence on the platform. On the other hand, if it can reach those who would have given up purchasing due to complex procedures, then its value is truly realized.
The Next Competition in Agent Commerce
This time, with the launch in Canada, "agent commerce" has been transformed from a concept into a concrete checkout action. The focus of competition will shift from who can generate recommendations that are more human-like to who can obtain reliable product data, who can handle payments and permissions, and who can provide traceable records in case of errors. For users, what is most important is not whether AI can press the purchase button on their behalf, but whether each step is transparent, reversible, and verifiable.
When search, assistance, and checkout are integrated into the same interface, the platform no longer merely acts as a traffic distributor but begins to function as a transaction coordinator. Retailers need to find a new balance between acquiring new conversions and maintaining customer relationships; consumers, on their part, must learn to distinguish between "system recommendations," "merchant commitments," and "actions authorized by themselves." AI Paying on behalf of others is just the first step; what truly determines whether this model can scale is whether trust and responsibility can flow as smoothly as payment does.
