Key statistics
- Agentic shoppers could account for $190 billion to $385 billion of US ecommerce spending by 2030, a 10% to 20% share. Morgan Stanley
- Only 9% of consumers are open to an AI agent making fully autonomous purchases, while 74% would trust an agent with routine tasks. Accenture
- Google says its Universal Commerce Protocol already enables direct checkout for hundreds of thousands of brands and retailers across Google. Google
- Walmart said conversion rates were three times lower for items sold directly inside ChatGPT's Instant Checkout than for items that sent shoppers to its own site. Modern Retail
Imagine telling an assistant, "Get the back-to-school shopping done and keep it under $400," and coming back to find everything ordered. That's the promise of agentic commerce: AI that doesn't just help you shop, but shops for you.
The idea moved quickly from theory to product in 2025 and 2026. OpenAI, Google, Amazon, Microsoft, Meta and the major payment networks have all launched agentic shopping features, and new standards now let AI agents check out with online stores.
The reality is more gradual than the headlines. Below, we'll explain what agentic commerce is, how it works, what's actually live today and what online brands should do about it.
What is agentic commerce?
Agentic commerce is shopping where an AI agent acts on a person's behalf: finding products, comparing options and completing the purchase within limits the person has set. The agent might be a general assistant like ChatGPT or Gemini, a retailer's own assistant, or a feature built into a browser or payment app.
The key word is "acts". In most AI shopping today, the AI researches and recommends, and the person still does the buying. In agentic commerce, the agent can build the cart, check out and pay.
The store usually stays the merchant of record. It takes the payment, ships the order, and handles returns and customer service, even though the customer never visited its website.
Agentic commerce vs. AI-assisted shopping
The difference comes down to who makes the final move:
| AI-assisted shopping | Agentic commerce | |
|---|---|---|
| What the AI does | Searches, compares and recommends | Searches, chooses, checks out and pays |
| Who completes the purchase | The shopper, usually on the store's website | The agent, within limits the shopper sets |
| Example | Asking ChatGPT to compare running shoes, then buying on the brand's site | Asking an agent to reorder your usual coffee when it's on sale |
| How common it is today | Widespread | Early, and mostly with the shopper confirming each purchase |
Most of what's happening today sits in between. Agents increasingly handle checkout, but the shopper confirms before anything is bought.
The levels of agentic commerce
Several companies have described agentic commerce as a series of levels. Stripe's 2025 annual letter sets out five:
| Level | What happens |
|---|---|
| 1. Eliminating web forms | You choose what to buy, and the agent fills in your payment and shipping details |
| 2. Descriptive search | You describe a situation, such as back-to-school shopping for a third grader, and the agent finds suitable products |
| 3. Persistence | The agent remembers your preferences, so its options already fit your taste and budget |
| 4. Delegation | You set the goal and budget, and the agent chooses and buys |
| 5. Anticipation | The agent buys what you need before you ask |
Stripe says level 4 is what most people mean when they talk about agentic commerce, and that the industry is "hovering on the edge of levels 1 and 2". McKinsey's automation curve makes a similar progression, from agents that assist shoppers to agents that act on standing goals and eventually negotiate with other agents.
Consumers move more slowly than the technology. In Accenture's 2026 survey of 25,590 people, willingness dropped sharply as agents took on more of the decision:
Consumers willing to hand shopping tasks to an AI agent, 2026
How agentic commerce works
An agentic purchase typically has five steps.
1. Discovery. The agent needs accurate product data. Stores share it through product feeds, such as Google Merchant Center, OpenAI's merchant feeds and Shopify's Catalog, which tell agents what you sell, what it costs and whether it's in stock.
2. Decision. The agent compares options against the shopper's request, using product details, reviews, prices and delivery times. Stores with complete, specific product information are easier for an agent to recommend.
3. Checkout. The agent creates an order with the store through a standard protocol, so it doesn't have to fill in a website's checkout form. Google's Universal Commerce Protocol and OpenAI and Stripe's Agentic Commerce Protocol are the two main standards.
4. Payment. Instead of handing over a card number, the agent passes a limited-use payment token. Stripe's shared payment tokens, for example, give a seller scoped access to the customer's payment method, with usage and expiry limits. Card networks have built their own versions: Mastercard Agent Pay uses agentic tokens, and Visa has Visa Intelligent Commerce.
5. Fulfillment and support. The store ships the order and sends updates back to the agent. Returns and customer service still run through the store.
Who's doing agentic commerce? (October 2026)
Here's where the major players stand:
| Company | What's available | Status |
|---|---|---|
| OpenAI (ChatGPT) | Product discovery from merchant feeds, with checkout increasingly on merchants' own sites | Live, after changes in March 2026 |
| Checkout in AI Mode in Search and the Gemini app through the Universal Commerce Protocol | Live in the US, expanding to more countries | |
| Microsoft | Copilot Checkout, where the merchant stays merchant of record | Rolling out in the US since January 2026 |
| Amazon | Alexa for Shopping, which combines Rufus and Alexa+, with price alerts and Auto-Buy, plus Buy for Me for products on other brands' sites | Live in the US, with Buy for Me in beta |
| Meta | Muse, a personal AI agent that checks out with Stripe's Link and asks for approval before buying | Launched in September 2026 |
| Shopify | Agentic Storefronts, which connect stores to AI channels including Google, ChatGPT and Copilot | Live, with direct checkout on by default for eligible US stores on Google |
| Visa and Mastercard | Tokenized payment programs for AI agents | Live, with Visa reporting hundreds of completed agent transactions in December 2025 |
Google's earlier version also deserves a mention. In November 2025, it launched agentic checkout in Search, which tracks a product's price and, once the shopper confirms, buys it on the merchant's site with Google Pay.
What happened to ChatGPT's Instant Checkout?
ChatGPT's Instant Checkout was the first big test of agentic commerce, and its story is instructive.
OpenAI launched it in September 2025, letting US users buy from Etsy sellers inside the chat, with more than a million Shopify merchants announced as coming soon. It supported single-item purchases, and merchants paid a small fee on completed sales.
It didn't sell as well as hoped. Walmart told Wired that conversion rates were three times lower for items sold directly inside ChatGPT than for items that sent shoppers to Walmart's own site.
In March 2026, OpenAI said the first version of Instant Checkout "did not offer the level of flexibility that we aspire to provide". It began letting merchants use their own checkout experiences and focused on product discovery instead.
The lesson for brands is that shoppers still want control at the moment of purchase. Agents are already changing how people find products, faster than they're changing how people pay.
How to get your store ready for agentic commerce
You don't need to build anything yourself to start. Most of the work is making sure agents can understand and trust your store.
Get your products into AI channels. On Shopify, Agentic Storefronts connects your catalog to supported AI channels. On other platforms, Google Merchant Center is the main route to Google's AI shopping, and OpenAI currently accepts product feeds for ChatGPT from approved partners.
Make your product data complete. Agents compare products on specifics: materials, dimensions, compatibility, sizing and use cases. Fill in every attribute your feeds support and use accurate identifiers such as GTINs.
Keep prices and stock accurate. An agent that tries to buy something out of stock creates a failed order. Stripe recommends keeping the share of agent checkouts that fail because an item is unavailable below 5%.
Publish clear policies. Agents and the platforms behind them need your return policy, shipping terms and customer service contacts, and shoppers rely on them more when they never see your website.
Decide which products to include. Some products, such as subscriptions and personalized items, aren't eligible for some AI checkouts yet. Check each channel's rules and your platform's settings.
Measure AI traffic separately. Google Analytics 4 now has an AI Assistant channel for visits from tools such as ChatGPT, Gemini, Copilot and Grok. Purchases completed inside AI checkouts may not appear in your web analytics at all, so check your platform's order reports too.
What agentic commerce means for customer retention
Agentic commerce could change who owns the customer relationship.
When an agent chooses the product, the brand has less chance to make an impression. And when the purchase happens inside Google or ChatGPT, you may get the order without the email subscriber, the site visit or the chance to recommend something else.
Brands can protect the relationship in a few ways:
- Link loyalty to AI channels. Google's protocol supports identity linking, so shoppers who connect their accounts get loyalty benefits when an agent buys from you.
- Make the post-purchase experience count. Delivery, packaging and the first support interaction are often your first real contact with an agent-acquired customer.
- Give customers a reason to come to you directly. Your app, loyalty program, exclusive products and member pricing all encourage customers to buy from you rather than through an agent.
- Become the default reorder. As agents move toward anticipating needs, the brands a customer already buys from regularly are the ones most likely to be reordered automatically.
Risks and open questions
Who's responsible when an agent gets it wrong? Liability between shoppers, agent developers and merchants is still being worked out. The UK's competition regulator has said businesses are responsible for what an AI agent does on their behalf, "in the same way you are responsible for what an employee does". That guidance applies to agents a business deploys itself, rather than to shoppers' own agents.
Fraud. Stores need to tell legitimate shopping agents apart from malicious bots. The card networks' tokens and new agent identity standards are designed to help.
Trust. Only 23% of US consumers trust generative AI to handle payments on their behalf, according to a Visa survey, even though 72% have used an AI assistant.
Competing standards. Google's UCP and OpenAI and Stripe's ACP do similar jobs. Ecommerce platforms such as Shopify already connect stores to both Google's and OpenAI's shopping channels, so most merchants won't have to choose.
Final thoughts
Agentic commerce is real, but it's earlier than the hype suggests. AI agents are already changing how people discover products, and checkout inside AI tools is growing, with most purchases still confirmed by the shopper.
For online brands, the preparation is practical and low-risk: accurate product data, clear policies, reliable stock and a post-purchase experience that turns an agent-placed order into a lasting customer. For the latest numbers on adoption and forecasts, see our AI in ecommerce statistics, and for the details of Google's standard, read our guide to the Universal Commerce Protocol.
