Agentic Checkout, Explained

What if asking was all it took?

Your cousin’s wedding in Charleston is next weekend, and you still have nothing to wear.

Normally, that would mean spending your evening bouncing between tabs, filtering dresses, checking size charts, and figuring out which ones will actually arrive before your flight. But what if you could skip all of that?

Instead, you tell your AI assistant exactly what you’re looking for: a chartreuse cocktail dress under $200, in your size, appropriate for an outdoor wedding, and delivered by Friday.

Within moments, you have a handful of options that check every box. You pick your favorite and ask the assistant to buy it. Rather than sending you to the retailer’s website to complete the purchase, the assistant handles the checkout for you. A moment later, there’s an order confirmation waiting for you.

That is agentic checkout in a nutshell.

Agentic checkout is when an AI agent completes a purchase on a shopper’s behalf. From managing the cart and calculating shipping and taxes to handling payment and confirmation, the agent can take care of the transaction without requiring the shopper to visit the retailer’s website. The retailer still processes the order and remains the merchant of record.

For shoppers, the experience feels effortless. Behind the scenes, though, a lot has to happen to make it possible.

AI agents need to communicate with retailers, access accurate product information, and securely complete transactions. That’s where protocols like the Universal Commerce Protocol, Agentic Commerce Protocol, and Agent Payments Protocol come in. They establish how these systems work together to turn a simple request into a completed purchase.

This guide breaks down how agentic checkout works, the technology making it possible, where it’s available today, and what businesses need to know as shopping continues to evolve.

What is agentic checkout?

The easiest way to understand agentic checkout is to look at where it begins and ends.

Think back to that chartreuse dress. Asking an AI assistant to find options that match your budget, size, style, and delivery deadline is part of the shopping journey. Once you choose a dress and authorize the AI agent to complete the transaction, you have moved into AI agent checkout.

That narrower scope is central to the agentic checkout definition. It refers specifically to executing the purchase, including creating the cart, calculating shipping and taxes, authorizing payment, and confirming the order.

Agentic checkout vs. agentic commerce.

Agentic commerce is the broader experience. It can cover everything from product discovery and comparison to purchasing and post purchase support.

Agentic checkout is one part of that journey: the transaction itself.

For example, finding and comparing five chartreuse dresses is agentic commerce. Authorizing an AI agent to purchase your favorite is agentic checkout.

Who stays the merchant of record?

Giving an AI agent permission to complete checkout does not make the agent the seller.

The retailer remains the merchant of record in current protocol models. It processes the payment and remains responsible for fulfillment, returns, refunds, and customer support.

So when that dress arrives at your door, the retailer behind it is still responsible for the order. How much of the customer relationship it keeps depends on the surface. The AI agent simply helped you get there.

How agentic checkout works

From finding the perfect dress to receiving an order confirmation, several steps happen behind the scenes to turn a shopper’s request into a completed purchase.

Here’s how the full journey unfolds, including where agentic checkout comes into play:

1. Intent capture: The process starts with the shopper explaining what they need. In our example, that is a chartreuse cocktail dress under $200, in the right size, appropriate for an outdoor wedding, and delivered by Friday. The agent interprets those requirements and turns them into criteria it can use to shop.

2. Product selection: Next, the agent searches available product information for options that meet those criteria. It can evaluate details such as price, size, availability, and delivery timing. Instead of making the shopper sort through every possible dress, the agent can narrow the choices to relevant options. This is where ecommerce AI agents move from understanding intent to taking action.

3. Checkout session: Once the shopper chooses a dress, the agent moves from product discovery to checkout, opening a session with the retailer’s system. The retailer calculates the final price, taxes, shipping options, and other order details. The shopper can then review and confirm the purchase before the agent proceeds.

4. Payment authorization and execution: After confirmation, the agent needs permission to pay without handling raw card information. A tokenized payment can provide a delegated, one time payment credential limited to a specific retailer and amount. Other models use a signed mandate that records what the shopper authorized. The retailer then processes the payment through its own payment provider.

5. Post-purchase handoff: Once payment succeeds, the retailer confirms the order and takes over fulfillment and customer support. Your dress is officially ordered and on its way to you, hopefully in time for your cousin’s wedding in Charleston. The retailer handles shipping, returns, refunds, and any other order needs.

The exact technical steps vary by protocol, but the basic principle stays consistent: the shopper authorizes the purchase, the agent facilitates it, and the retailer completes the transaction.

Agentic checkout vs. traditional and one-click checkout

The biggest difference between agentic checkout vs. traditional checkout is who does the work.

With traditional checkout, the shopper moves through the retailer’s website and completes each step. One click checkout speeds up that process by using saved information, but the shopper still initiates and completes the purchase. With agentic checkout, an AI agent can handle much of that process on the shopper’s behalf after receiving authorization.

Think about our chartreuse dress. With traditional checkout, you find the dress, add it to your cart, enter your information, select shipping, and pay. One click checkout removes some of those steps. Agentic checkout goes further by allowing the AI agent that helped find the dress to complete the transaction for you.

That does not make traditional checkout optimization obsolete. Businesses still need a fast, intuitive checkout for shoppers who visit their storefront directly.

Agentic checkout simply introduces another path to purchase, one designed for situations where an AI agent is doing more of the shopping on the customer’s behalf.

The protocol landscape: UCP, ACP, and AP2

The alphabet soup of agentic commerce protocols can make agentic checkout sound more complicated than it needs to be.

The key is understanding that these protocols do not all solve the same problem. The Universal Commerce Protocol (UCP) and Agentic Commerce Protocol (ACP) operate primarily at the commerce layer, helping AI agents interact with businesses and their checkout systems. The Agent Payments Protocol (AP2) focuses on payment authorization and proving that an agent has permission to act for the shopper.

In other words:

• UCP and ACP help an agent shop and transact with a business

• AP2 helps establish what the shopper authorized the agent to do

• UCP vs. ACP is the more direct comparison because their commerce and checkout capabilities overlap

• No single protocol has become the universal standard

Universal Commerce Protocol (UCP).

The Universal Commerce Protocol (UCP) is an open source standard developed by Google in collaboration with companies including Etsy, Wayfair, Target, Shopify, and Walmart. Google announced it in January 2026.

UCP establishes a common language for AI agents, businesses, and payment providers across the commerce journey. Designed to work with existing retail infrastructure, UCP can connect with AP2 for payment authorization and communication protocols such as the Model Context Protocol (MCP), which helps AI agents access external tools and data, and the Agent2Agent Protocol (A2A), which allows different AI agents to communicate and coordinate tasks.

For example, once you’ve found that chartreuse dress for your cousin’s wedding, UCP could help your AI assistant communicate with the retailer’s systems to complete the checkout on your behalf.

Learn more about the Universal Commerce Protocol.

Agentic Payments Protocol (AP2).

The Agent Payments Protocol (AP2) addresses a different question: How can a business verify that a shopper actually authorized an AI agent to make a purchase?

AP2 uses cryptographically signed mandates to provide verifiable proof of what a shopper has authorized. In April 2026, Google contributed AP2 to the FIDO Alliance to support the development of open industry standards for secure payments made through AI agents.

How the protocols fit together.

Think of these protocols as layers rather than three competing ways to do the same thing.

UCP and ACP can tell an AI agent how to interact with a business and its checkout. AP2 can provide a way to establish what the shopper authorized the agent to purchase. Lower in the stack, protocols such as MCP and A2A can help systems and agents exchange information.

Card networks are also helping make agentic checkout more secure. Visa’s Trusted Agent Protocol helps retailers verify legitimate AI agents, while Mastercard Agent Pay focuses on authenticating agents and securing authorized payments. Both complement protocols like UCP and ACP rather than replacing them.

For businesses, the takeaway is simple: there is not one protocol to rule them all. The important question is which layer a protocol addresses and which shopping experiences a business wants to support.

Where agentic checkout is live today

So, where can you actually use agentic checkout?

The answer is already changing quickly. Some AI assistants can take shoppers all the way through checkout today. Others limit the experience to certain businesses, products, or approved integrations.

Status as of October 2026: Here is where AI assistant checkout stands across some of the major shopping experiences.

Google AI Mode and Gemini.

Shoppers can encounter agentic checkout in Google AI Mode and Gemini, with availability expanding through the Universal Commerce Protocol.

For participating U.S. retailers, shoppers can discover products and complete eligible purchases using Google Pay without leaving Google’s shopping experience. The retailer remains the merchant of record.

In our dress example, that could mean asking for a chartreuse cocktail dress, comparing the options Google surfaces, and purchasing an eligible dress within the same experience.

Perplexity.

Perplexity checkout is live in the U.S., with PayPal powering purchases for eligible products.

The experience is designed to keep the shopper in the conversation as they move from research to purchase. When checkout is available, shoppers can use PayPal to complete the transaction rather than leaving Perplexity and starting over on another site.

Microsoft Copilot.

Copilot Checkout has been available in the U.S. since January 2026.

A shopper can discover an eligible product through Copilot and complete the purchase within the experience. PayPal supports Copilot Checkout alongside Shopify and Stripe, while the retailer remains the merchant of record.

The important distinction is that Copilot helps facilitate the purchase. The retailer still owns the order and fulfillment, though how much of the customer relationship it keeps depends on the experience.

Amazon Buy for Me.

Amazon takes a different approach with Buy for Me.

When Amazon finds an eligible product sold on another brand’s website, the shopper can ask Amazon to purchase it on their behalf. Amazon’s agent then visits the external site and completes the order using the shopper’s encrypted payment and shipping information.

The brand fulfills the purchase and handles returns, exchanges, and customer service.

Unlike the other experiences, Amazon’s Buy for Me uses its own agentic shopping technology to complete purchases on external brand websites.

ChatGPT.

ChatGPT helped put agentic checkout on the map, but its approach has changed.

OpenAI introduced ChatGPT Instant Checkout with Stripe and the Agentic Commerce Protocol in September 2025. In March 2026, OpenAI said Instant Checkout is moving to apps, scaling back plans for checkout directly from shopping results and shifting its focus toward purchases through connected merchant apps. Because this is changing quickly, businesses should confirm current availability before planning around it.

That evolution is a useful reminder of just how early this market is.

A brief history: From the 2025 pilots to today

Agentic checkout moved from an emerging idea to a real way to buy in a remarkably short time.

The first wave brought payment frameworks, new agentic commerce protocols, and checkout experiences directly into AI assistants. Some of those early models have already changed course, which is why agentic checkout history is useful context for understanding where the technology stands today.

Here are the major milestones so far:

1. April 2025: Mastercard announced Agent Pay, introducing technology designed to support trusted payments initiated through AI agents.

2. September 2025: Google announced the Agent Payments Protocol, or AP2, to establish a common framework for securely authorizing agent led payments.

3. September 29, 2025: OpenAI and Stripe introduced the Agentic Commerce Protocol alongside ChatGPT Instant Checkout, allowing eligible shoppers to purchase products within ChatGPT.

4. October 2025: PayPal announced Agentic Commerce Services as part of its push into PayPal agentic commerce, with Commerce among its initial pilot partners.

5. November 25, 2025: Perplexity and PayPal launched in chat shopping and checkout in the U.S., allowing eligible shoppers to move from product discovery to purchase without leaving Perplexity.

6. January 2026: Google introduced the Universal Commerce Protocol at NRF. That same month, Microsoft launched Copilot Checkout, adding another major AI assistant to the agentic shopping landscape.

7. March 2026: OpenAI scaled back its original ChatGPT Instant Checkout approach, shifting its focus from purchases directly within shopping results to checkout through connected merchant apps.

8. April 2026: Google contributed AP2 to the FIDO Alliance, moving work on agent payment authorization toward broader, neutral industry governance.

That short timeline also reveals something important: agentic checkout is real, but the way it reaches shoppers is still evolving. The early years have been less about one model winning and more about businesses, AI companies, and payment providers testing different ways to connect discovery, authorization, and purchase.

How a merchant thinks about picking a protocol

There is no single answer to which agentic commerce protocol a business should support.

The better starting point is to look at how your customers shop, what your existing technology supports, and what needs to happen for an AI agent to complete a purchase successfully.

In short:

• Start with the AI assistants your customers actually use

• Check what your existing ecommerce and payment stack already supports

• Treat ecommerce and payment authorization as separate decisions

From there, businesses can work through four practical questions.

1. Start with the surfaces your customers already use.

Before debating UCP or ACP, look at where your customers are actually shopping with AI.

If they are using Google AI Mode or Gemini, UCP may be more relevant. If they are shopping through experiences built around ACP, that protocol becomes more important.

Think about our wedding guest. She does not care which protocol helps her buy the chartreuse dress. She cares that the assistant she already uses can find it and complete the purchase.

For businesses evaluating agentic checkout for merchants, customer behavior is a more useful starting point than the protocol itself.

2. Check what your platform and payments provider already support.

Most businesses will not need to build every protocol integration themselves.

Check whether your ecommerce platform or payments provider has published support for the protocol you need and whether that support is generally available, beta, or early access. Your existing stack may already expose some of the capabilities required for agentic checkout.

This is also where evaluating agentic commerce platforms can help clarify what your current technology makes available to AI agents.

3. Decide the commerce layer first, the payment layer second.

UCP and ACP primarily address how agents interact with commerce systems, while AP2 addresses payment authorization.

That means choosing between UCP and ACP is a different decision from determining whether AP2 matters to your payment flow. AP2 can sit underneath the commerce experience rather than replacing it.

Start with the shopping surfaces you need to support. Then determine what payment authorization those experiences require.

4. Keep your catalog and checkout data clean regardless of protocol.

Protocols may change, but AI agents still need accurate information.

Product names, descriptions, prices, inventory, variants, shipping details, and checkout data should be structured and current. Your systems also need to expose enough information for an authorized agent to create and complete a checkout session.

That work pays off regardless of whether UCP, ACP, or another approach becomes more important to your business.

There are also signs that it may be too early to enable a protocol:

• Your customers are not showing demand for AI assistant checkout

• Your platform or payment provider has not released support

• Your product and checkout data are not ready for agents to use reliably

Waiting in those cases does not mean ignoring agentic checkout. It means getting the fundamentals in place while watching where your customers and technology partners go next.

Preparing your store for agentic checkout

Preparing for agentic checkout starts long before an AI agent tries to place an order.

Agents need accurate product information, checkout systems they can interact with, secure ways to authorize payment, and clear rules for what they can do. Building that foundation now can make it easier to support new protocols and AI shopping experiences as they become relevant to your customers.

Audit your checkout flow for automation readiness.

Start with the checkout experience you already have.

Look for manual steps, disconnected systems, or other friction that could prevent an authorized agent from completing a transaction. An AI-ready checkout should allow an agent to create a checkout session, retrieve accurate totals, select available shipping options, and submit an authorized order through clearly defined systems.

This work also benefits shoppers completing purchases themselves. Improving checkout optimization can create a stronger foundation for both human and agent led transactions.

Publish structured product data and clear API endpoints.

An AI agent is only as useful as the information it can access.

Your structured product data should clearly communicate details such as price, inventory, variants, product attributes, and fulfillment options. In our dress example, the agent needs to know that the chartreuse dress is available in the shopper’s size, costs less than $200, and can arrive before Friday.

Businesses should also provide machine-readable feeds, manifests, and APIs that let authorized agents access product information and interact with checkout. An API-first ecommerce approach can make these interactions easier to support across different shopping experiences.

Choose payment providers that support tokenized agent payments.

Agentic checkout should not require an AI agent to handle a shopper’s raw card details.

Look for payment gateways and providers that support secure agent transactions through approaches such as delegated payment tokens or signed mandates. These methods can limit what an agent is authorized to purchase while allowing the retailer to process the payment through its existing provider.

Support will vary by provider and protocol, so confirm what is currently available and whether the capability is generally available, beta, or early access.

Build trust with transparent AI policies and opt in settings.

Shoppers should know when an AI agent is acting for them and what they are authorizing it to do.

Clearly explain what information an agent can access, how that information is used, and when the shopper must approve an action. Give customers meaningful controls over agent purchases and make policies around cancellations, returns, refunds, and disputes easy to find.

The goal is not simply to automate more of checkout. It is to make that automation understandable and trustworthy.

Agentic checkout readiness checklist.

Before enabling agentic checkout, make sure you can:

• Keep product prices and inventory current

• Provide complete, structured product attributes

• Make fulfillment options machine readable

• Create and complete checkout sessions through APIs

• Support secure tokenized or mandate based payments

• Define what AI agents can and cannot do

• Give shoppers clear authorization and control

Benefits and trade offs of agentic checkout

Agentic checkout can make buying easier, but easier does not automatically mean better for every shopper or business.

The opportunity is real. So are the unanswered questions around customer relationships, consent, fraud, and agentic checkout conversion. For now, businesses should consider both sides.

For businesses.

The biggest benefits of agentic checkout come from meeting customers where more shopping decisions are beginning to happen.

• A shorter path to purchase: An agent can move from product discovery to checkout without asking the shopper to restart the journey on another site.

• Access to new shopping surfaces: Businesses can make products available to customers shopping through AI assistants, not just traditional search engines and storefronts.

• More purchase-ready traffic: Adobe Digital Insights reported in April 2026 that AI referred traffic to U.S. retail sites grew 393% year over year in the first quarter. In March, those visits converted 42% better than non AI traffic. These figures measure AI referred traffic, not agentic checkout specifically, but they show how AI is influencing shoppers closer to purchase.

For shoppers.

Agentic checkout can remove some of the tedious work between finding the right product and actually owning it.

Think about our chartreuse dress. Instead of finding the perfect option and then starting a separate checkout process, the shopper can authorize the agent to keep going.

That can mean fewer forms to complete, fewer screens to navigate, and a more continuous path from “this is the one” to “order confirmed.”

Trade offs and open questions.

The agentic checkout risks matter just as much as the convenience.

• Fewer direct customer touchpoints: When checkout happens somewhere else, businesses may lose opportunities to deliver their own brand experience and collect first party insights.

• Dependence on AI assistants: Businesses have less control over how products are presented when another platform owns the shopping interface.

• Consent and fraud: Businesses need reliable ways to know what a shopper authorized an agent to purchase.

• Uncertain adoption: Early results show that removing checkout steps does not guarantee higher conversion. Walmart reported in March 2026 that purchases completed within ChatGPT converted at one third the rate of transactions that clicked through to Walmart’s website.

That last point is important. Agentic checkout may reduce friction, but shoppers still value context, trust, and control when they are ready to buy. The opportunity is not simply to automate more. It is to determine where automation genuinely makes the shopping experience better.

The final word

Agentic checkout is real, but it is still early. UCP, ACP, and AP2 are helping define how AI agents, businesses, and payment providers work together, but they address different layers of the transaction rather than competing to become one universal standard.

For businesses, the next steps are practical:

1. Get your data ready. Structure product and checkout data so AI agents can accurately understand what you sell and how to buy it.

2. Check your existing technology. Find out which protocols your commerce platform and payments provider support and whether that support is generally available, beta, or early access.

3. Follow your customers. Pay attention to the AI assistants they actually use and prioritize the shopping experiences that matter to them.

Agentic checkout may change where checkout happens, but the goal remains familiar: help customers get from “I want that” to “it’s on the way” with as little friction as possible.

For a deeper look at the technology behind agentic checkout, explore our guides to the Universal Commerce Protocol and Agentic Commerce Protocol.

FAQs about agentic checkout