Walmart Opens Shopping to AI Agents: ChatGPT and Gemini Can Place Orders

Walmart Opens Shopping to AI Agents: ChatGPT and Gemini Can Place Orders

Walmart is handing over the shopping cart to AI agents. On October 21, Walmart announced the launch of Agent Profiles in Agent Gateway, allowing AI agents such as ChatGPT, Gemini, and Meta's Muse to complete shopping on behalf of users. In contrast to Amazon's active blocking of AI shopping agents, Walmart has chosen an open approach.

How Agent Profiles Work

The core mechanism is "one agent, one profile." Each AI agent has an independent identity in Walmart's system, and users can set spending limits for that profile or require manual approval for every transaction. Orders generated by agents are kept separate from the user's main account, permissions can be revoked at any time, and agents can never access the user's bank card data.

This design keeps control in the user's hands: users enjoy the convenience of automated shopping without worrying about AI "slip-ups" or unauthorized spending. Walmart specifically notes that using Agent Profiles does not require a Walmart Plus subscription, lowering the barrier to entry.

Agent Gateway can be understood as a unified access layer that the retailer has established for AI agents: unlike the website or app interface for human users, it provides a set of machine-readable interfaces and identity management mechanisms responsible for agent registration, authentication, permission verification, and transaction auditing. Agent Profiles complete identity registration and policy binding at this layer, and every call made by an agent must pass the gateway's rule checks. The advantage of this architecture is that the platform can centrally enforce limits, approvals, and revocation policies at a single entry point, without needing to customize integration logic for each AI vendor separately, while also leaving room for expanding to more agent partners in the future.

Divergence from Amazon's Approach

According to analysis from the video source, Amazon is actively blocking AI shopping agents from accessing its platform, while Walmart has taken the completely opposite stance. This divergence reflects the retail giants' different judgments about AI agents: one side sees them as a traffic entry point, while the other worries that agents bypass the platform's advertising and recommendation systems, weakening its influence over consumer decisions.

Walmart's open strategy is more about seizing the ecological niche of "agent-friendly retailer"—once users are accustomed to letting AI shop for them, the platform that supports agents first may gain a default-option advantage.

Understanding the divergence between the two giants requires returning to how AI shopping agents crawl and interact. Currently, most agents either simulate human operations through web scraping and browser automation, or complete orders through platform-open APIs. The former bypasses the platform's recommendation slots, ad placements, and personalized pricing logic, rendering the platform's guidance of consumer decisions ineffective, while also bringing server load and bulk data scraping issues; the latter allows the platform to retain control over product ranking, promotional exposure, and user data. Amazon's blocking of agents stems from its core concern that the commercial value of its advertising and recommendation systems will be diluted; Walmart's unified access through Agent Gateway, by contrast, brings agent traffic under its own rules, trading structured authorization for an early ecological niche advantage.

Security Boundaries and Rollout Pace

The security design of Agent Profiles deserves attention: order isolation means users can clearly distinguish which purchases come from AI and which from themselves; the immediacy of permission revocation provides a means of cutting losses. The provision of not touching card data essentially replaces the exposure of raw payment credentials with tokenized authorization.

According to the original video, Agent Profiles will be rolled out gradually "next month," with specific coverage and the agent access list not yet fully disclosed. The appearance of Meta's Muse as one of the access partners also indicates that Walmart has adopted a multi-partner strategy in its agent ecosystem rather than binding itself to a single AI vendor.

Tokenization is a standard security mechanism in the payment card industry: the user's real card number is not directly given to the merchant or agent; instead, the payment network or token service provider generates a substitute "token" that can only be used at a specific merchant, in a specific scenario, or on a specific device, and can be revoked at any time. Walmart's statement that "agents can never access bank card data" relies precisely on this layer of abstraction—agents receive restricted credentials rather than raw card numbers that could be misused. This also means that even if an agent identity is stolen, the damage an attacker can cause is limited to the preset token permissions, which is incomparable to the risk of leaking a real card number.

Impact on the AI Shopping Ecosystem

Agent Profiles are essentially an authorization and auditing framework, not simply an API opening. It answers a key question: when AI executes transactions on behalf of humans, how are responsibility boundaries divided? Walmart's answer is—agents have independent identities, independent limits, and independent orders, while humans retain the final veto power.

If this model proves viable, it could become a reference template for the retail industry to integrate AI agents. For consumers, the real test lies in approval fatigue: if every transaction requires confirmation, the convenience of automation will be greatly reduced; if spending limits are set too broadly, the risk of mistaken purchases may be amplified. Whether Walmart can find a balance between the two will determine whether agent shopping is a practical tool or a demo concept.

One-line Summary

Walmart launches Agent Profiles, allowing AI agents to shop for users, taking an open approach against Amazon's blocking.

Paragraph Summary

Walmart has launched Agent Profiles in Agent Gateway, allowing AI agents such as ChatGPT, Gemini, and Meta Muse to shop on behalf of users. The core mechanism is "one agent, one profile": each agent has an independent identity, users can set spending limits or require per-transaction approval, agent orders are isolated from the main account, permissions can be revoked at any time, agents cannot access bank card data, and no Walmart Plus subscription is required. In contrast to Amazon's blocking of AI shopping agents, Walmart has chosen open access, aiming to seize the ecological niche of "agent-friendly retailer." Agent Profiles are essentially an authorization and auditing framework that clarifies responsibility boundaries for AI shopping, but the balance between approval fatigue and limit strictness will determine its practical value. The feature will roll out gradually next month.