Yes, some AI agents can search products, build carts, and—where a retailer or platform enables it—complete checkout. But there is no universal permission system. Retailers are deciding how to distinguish an authorized shopping agent from a scraper or fraud bot, what data an agent may access, when a shopper must approve an action, and whether the merchant remains responsible as seller of record.
Can AI agents shop on retail websites?
An AI shopping agent is software that can interpret a shopper’s goal and perform multiple steps, rather than merely returning search links. Depending on its permissions and the retailer’s integration, it might compare products, check availability, apply a discount code, place an item in a cart, submit payment, or handle an order question.
As an Amazon Associate I earn from qualifying purchases.
Those capabilities are not interchangeable. A search crawler collecting public pages, a training crawler copying content, a conversational assistant recommending products, and an agent authorized to buy are different types of automated activity. A retailer may welcome one and restrict another.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchPC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11The practical question is therefore not whether “AI bots” are allowed. It is what an identified agent is allowed to do, for which customer, with what approval and accountability.
#1 Best Overall
Why retailers are cautious about opening access
Retailers have traditionally used bot-detection systems to limit scraping, inventory hoarding, credential attacks, payment fraud, and automated abuse. Those systems look for traffic patterns, device signals, login behavior, and transaction anomalies. A legitimate agent can resemble the very automation those controls were built to stop.
Visa described this classification problem in its October 14, 2025 protocol announcement: merchant systems can mistakenly block legitimate agentic transactions, while merchants still need to separate trusted agents from malicious automation. That is an industry problem statement, not evidence that every retailer blocks agents or follows one shared rule.
Retailers also have commercial reasons to be selective. An agent may send a shopper directly to a product page while bypassing brand messaging, loyalty prompts, or merchandising experiments. It may make many price and availability requests, increasing infrastructure costs. If an agent chooses the wrong size, substitutes an unsuitable grocery item, or submits an unwanted recurring order, the merchant must handle the complaint even if software made the decision.
Free tools Windows power users keep installed
One-click scans. No signup required.
The six questions behind an access decision
- Permission scope: Is the agent only reading product data, or can it access an account, construct a cart, submit payment, or request a return?
- Agent identity: Can the agent present a verifiable identity, its operator, and its declared commercial purpose?
- Consumer control: Which actions require explicit approval, and can a shopper reverse them?
- Data and payment access: What profile, order, loyalty, or payment information is disclosed, and for what stated purpose?
- Retailer relationship: Does the merchant retain the customer relationship, order visibility, and seller-of-record role?
- Abuse handling: Can the system detect credential theft, scraping, manipulated instructions, or abnormal purchasing without blocking ordinary customers?
How do retailers know whether an AI agent is legitimate?
Reliable identification has to be stronger than a browser user-agent string, which can be forged. Emerging designs combine cryptographic credentials with information about the agent’s purpose and the shopper’s relationship with the merchant.
Visa Trusted Agent Protocol
Visa’s October 2025 release describes agent-specific cryptographic signatures and three categories of information:
Rank #2
- Ideal for Gifting
- Ideal for a bookworm
- Compact for travelling
- Intent: whether the agent is retrieving product details or attempting to purchase a specified product.
- Consumer recognition: information indicating an existing account or prior interaction between the shopper and merchant.
- Optional payment data: payment information supplied only when the transaction requires it.
Visa says the initial specifications apply to its network in this phase. The company describes work with Cloudflare and feedback from payment and commerce companies, while coordinating with groups including the Internet Engineering Task Force, the OpenID Foundation, and EMVCo. That makes it an emerging network approach, not a universal standard that every retailer has implemented.
Visa’s Jack Forestell, Chief Product and Strategy Officer, said: “We believe the entire payments ecosystem has a responsibility to ensure sellers can trust AI agents as much as they trust their best customers and networks.” This is Visa’s position on the design goal, not an independent certification of current merchant systems.
Universal Commerce Protocol
Google announced the Universal Commerce Protocol (UCP) on January 11, 2026, describing it as an open standard for discovery, buying, and post-purchase support. Google identified Shopify, Etsy, Wayfair, Target, and Walmart as co-developers and said more than 20 additional ecosystem participants had endorsed it.
Google’s initial checkout description is narrower than a blanket web permission. It covers eligible U.S. retailers whose products appear in Google listings, with checkout in AI Mode in Search and in the Gemini app. Google says the retailer remains seller of record. It also announced Business Agent, Merchant Center attributes for conversational discovery, and a Direct Offers pilot. Eligibility and rollout depend on the participating integration.
Google’s Ashish Gupta, Vice President and General Manager of Merchant Shopping, said: “The shift to agentic commerce will require a shared language across the ecosystem—and the Universal Commerce Protocol provides that framework.” The statement describes Google’s rationale; it does not establish universal adoption.
Platform-level integrations
Shopify describes UCP integrations that let agents handle checkout details such as discount codes, loyalty credentials, subscription cadence, and selling terms. Its announcement also describes embedded checkout through Google AI Mode and Gemini, plus a Copilot Checkout integration. These are platform capabilities and integration descriptions, not proof that every merchant using the platform—or every shopper—has access to them.
| Approach | What it addresses | Scope and qualification |
|---|---|---|
| Visa Trusted Agent Protocol | Cryptographic agent identity, declared intent, consumer recognition, and optional payment data | Visa’s initial network specifications announced October 14, 2025; not a universal implementation |
| Universal Commerce Protocol | Discovery, checkout, and post-purchase interactions | Google-announced open standard with specified eligible U.S. checkout scenarios as of January 11, 2026 |
| Platform integrations | Operational details such as discounts, loyalty credentials, subscriptions, and selling terms | Availability depends on the merchant, platform, geography, and participating integration |
Can AI agents buy things for me?
They can in a supported flow, but “buy for me” should mean delegated execution under defined limits—not unlimited authority. A safer transaction separates product selection, approval, payment authorization, and fulfillment responsibility.
- Set the shopping objective. Specify constraints such as budget, acceptable brands, dimensions, delivery date, dietary rules, or return requirements. Vague instructions increase the chance of an unsuitable choice.
- Let the agent discover and compare. The agent retrieves product facts, prices, stock, shipping estimates, reviews, or eligible offers. The retailer may allow this read-only activity while restricting account or checkout access.
- Review the proposed item and terms. Check the exact model, quantity, seller, taxes, shipping, warranty, subscription cadence, and return conditions. Product titles and images can hide materially different variants.
- Approve the decision that matters. Require confirmation before an agent commits a purchase, changes a recurring order, uses loyalty points, or selects a substitute. A saved spending limit can govern routine orders, but it should be visible and revocable.
- Authorize payment through the supported checkout. Do not paste a full card number or account password into an unverified agent. A protocol or platform may pass a payment token or other limited authorization instead of exposing reusable credentials.
- Keep the receipt and control path. Confirm which merchant accepted the order, who is seller of record, where cancellation and returns are handled, and how to revoke the agent’s access.
Google says its initially described U.S. UCP checkout keeps the retailer as seller of record. That matters because the merchant remains the party responsible for fulfillment, refunds, taxes, and customer-service obligations in that flow; other integrations may structure responsibility differently.
What consumers are actually doing with AI assistants
Interest in assistance is broader than authorization to purchase. Visa’s 2025 research, covering the United States, Australia, and New Zealand, examined discovery through post-purchase behavior for birthday gifts, electronics, groceries, and travel. The samples were 1,600 people in the United States, 1,600 in Australia, and 500 in New Zealand. Visa reported that 55% of U.S. respondents were familiar with AI assistants; that measures familiarity, not completed agent purchases.
Visa also cited Adobe Data Insights for more than 4,700% growth in AI-driven traffic to U.S. retail websites over the preceding year. “Traffic” is not a count of purchases or unique shoppers. In the same Visa account, 85% of surveyed shoppers who had used AI to shop said it improved their experience; that percentage applies to the AI-using group described by Visa.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Rank #4
The National Retail Federation and IBM’s January 2026 summary reported separate behavior measures from IBM proprietary research involving 18,000 global consumers. Respondents said they used AI assistants to research products (41%), find reviews (33%), and search for deals (31%). Those are research and discovery activities, not autonomous-purchase rates. The summary also said nearly three-quarters still shop in physical stores.
Trust remains conditional. NRF and IBM reported that 52% were comfortable sharing their data, while 83% reported overlapping concerns involving privacy, misuse, and unwanted marketing; the 83% is not one single concern or a mutually exclusive response. Visa’s averages across its three markets found around 85% wanted explicit control over agent data access, nearly 43% worried an agent could choose the wrong product, and about 50% feared decisions could be made without them.
Why are retailers blocking AI bots?
Some automated requests are blocked because they look like scraping, fraud, denial-of-service activity, inventory manipulation, or account takeover. A retailer may rate-limit an unidentified crawler while allowing a signed agent that states its purpose and ties the request to a recognized customer.
The evidence available does not establish a retailer-wide rule across named merchants, nor does it show that retailers broadly block all AI agents. Industry workshops and reports discuss governance, but a summary of the National Retail Federation’s separate report with PwC’s Center for Digital Risk & Innovation says workshops in late 2025 involved U.S. retail cybersecurity, technology, legal, and business leaders without publishing a common access policy or detailed controls.
A practical permission model for retailers
Retailers can make policy easier to understand by assigning different controls to different actions instead of treating all automation alike.
Best Value
- It can be a gift option
- Comes with secure packaging
- Helpful in various ways
| Permission tier | Typical capability | Controls a retailer may require |
|---|---|---|
| Public discovery | Read product descriptions, prices, availability, and shipping information | Rate limits, accurate bot identification, cache rules, and monitoring for scraping abuse |
| Personalized shopping | Use a signed-in profile, loyalty status, saved preferences, or order history | Verified agent identity, customer authorization, purpose-limited data, and audit logs |
| Cart construction | Add items, apply offers, select variants, or propose substitutions | Clear item and price display, expiry checks, and shopper review before commitment |
| Checkout | Submit an order and payment authorization | Cryptographic identity, transaction limits, step-up confirmation, fraud screening, and explicit terms |
| Post-purchase service | Track delivery, cancel, return, or change a subscription | Strong account authentication, reversible actions where possible, and a human escalation path |
This tiered approach lets a merchant support useful discovery without granting an unknown program the power to spend money or change an account.
What shoppers should check before delegating a purchase
- Identity: Is the agent operated by a recognizable provider, and does the checkout show which merchant receives the order?
- Authorization: Can you set a spending cap, product rules, expiration time, and approval requirement?
- Data minimization: Does the agent need your full order history, loyalty balance, address book, or payment credentials for this task?
- Decision visibility: Will you see the exact variant, seller, taxes, shipping, recurring terms, and return policy before submission?
- Recovery: Can you revoke access, cancel an order, dispute a payment, or reach the retailer without going through the agent?
- Substitutions: Are replacements prohibited, pre-approved, or selected automatically? Grocery and travel decisions can carry higher consequences than a minor price difference.
What agentic commerce means for personal finances
Agentic commerce is the use of software agents to perform commerce tasks across discovery, purchase, and service. Its financial effect depends less on how conversational the interface feels than on the controls around it.
Automation can reduce missed discounts, comparison time, and routine reordering. It can also increase spending frequency, trigger subscriptions, combine incompatible promotions, or buy a plausible but unsuitable substitute. A household budget should therefore treat an agent’s spending authority like a recurring payment permission: define a ceiling, list excluded categories, require confirmation for exceptions, and review transaction alerts.
Recommended Free Tools
Retailers and payment networks are still deciding how identity, authorization, data sharing, and liability fit together. Protocol announcements from Visa and Google show competing or complementary ecosystem efforts, not a settled market rule. Until those roles are clear, the safest assumption is that the shopper must remain able to see, approve, and undo consequential decisions.
Where the market is heading
Near-term adoption is likely to be uneven. A retailer may expose product data to conversational discovery, support an embedded checkout for a limited geography, and still reject unsanctioned account automation. Another may permit a partner agent but not an independent browser extension. Geography, merchant systems, payment network, customer status, and product category can all change the result.
The key test for any announcement is operational: Which actions are covered? Which shoppers and countries qualify? What identity evidence is exchanged? Who stores or tokenizes payment data? What requires confirmation? Who is seller of record? How are fraud, returns, and disputes handled? Answers to those questions tell you far more than a claim that an agent “can shop.”
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
