Agentic commerce explained: How ChatGPT Is testing AI’s role in shopping
When OpenAI announced “Buy it in ChatGPT” this week, it wasn’t just unveiling a new convenience feature. It was taking a step toward something bigger: a world where AI agents don’t just talk, but act on our behalf. At the center of this experiment is a concept called Agentic Commerce – a way to let conversational AI models handle the messy middle of e-commerce, from product discovery to checkout.
SurveyAlso read: ChatGPT now lets you buy products directly through chat with Instant Checkout: How it works
What is agentic commerce?
Agentic commerce is the idea that AI agents should not just answer questions but also complete tasks. Instead of copying a product link or switching apps, a user could ask an AI: “Find me a handmade ceramic mug under $30 and order it.” The agent then searches, confirms the choice, processes payment, and sends the order – all within the chat interface.
OpenAI is testing this vision through a new feature called Instant Checkout, currently rolled out for U.S. users buying from Etsy sellers. The checkout flow happens entirely inside ChatGPT, powered by the Agentic Commerce Protocol, an open standard developed with Stripe.
How It Works
- Discovery inside ChatGPT: A user browses or asks about products.
- Agent-driven checkout: Once a choice is made, the AI initiates a step-by-step confirmation, ensuring the user stays in control.
- Secure payment handling: Instead of passing credit card details to the AI, ChatGPT uses tightly scoped tokens with Stripe.
- Merchant fulfillment: Sellers still handle shipping, support, and returns through their own systems.

The goal: ChatGPT doesn’t become an e-commerce platform itself. Instead, it becomes a universal interface, stitching together the user’s intent, payment rails, and merchant infrastructure.
Why this matters for AI agents
The significance isn’t just about shopping convenience. It’s about testing how far AI agents can go in bridging intent and action. Until now, most chatbots have been “assistants” – they can recommend, summarize, or guide. With agentic commerce, the AI is edging into the role of executor: carrying out transactions under real-world constraints like payments, inventory, and logistics.
This represents a shift from conversation to action, and it mirrors a broader industry trend: AI agents that can book flights, manage calendars, write code, or trade stocks. Commerce is simply one of the most tangible and high-stakes testbeds.
Guardrails and limitations
OpenAI emphasizes that the AI cannot buy things on its own. Every step requires explicit user approval. Sensitive data like payment details never pass through the model. These design choices are crucial, both for security and for building trust.
Still, there are limits:
- Only single-item purchases are supported today (no shopping carts yet).
- Only U.S. users and U.S. Etsy merchants are included in the pilot.
- Complex scenarios like bundles, customizations, or international shipping remain out of scope.

Also read: OpenAI plans to launch a social app for AI videos: Here’s how it may work
The experiment is narrow by design, but it’s meant to test whether users feel comfortable transacting through an AI layer.
What’s at stake
If agentic commerce scales, it could reshape how consumers interact with merchants. Instead of going to Amazon, Etsy, or Shopify directly, people could start in ChatGPT, turning the AI into a kind of meta-layer over the internet’s marketplaces. That has implications for merchants who gain access to new sales channels but risk losing direct relationships with customers, payment processors who may see AI as a new distribution layer for their rails, consumers who could experience shopping that is more conversational, personalized, and frictionless and AI governance since letting models handle money and orders raises new questions of accountability, liability, and consumer protection.
Agentic commerce is an early glimpse of what AI-powered transactions might look like across industries. Imagine asking an AI to plan a vacation and book the flights and hotels, order recurring groceries from multiple vendors or negotiate bills or subscriptions.
Each of these requires the AI to not just inform but act in structured, reliable ways, which is exactly what protocols like the Agentic Commerce Protocol aim to enable.
For now, ChatGPT’s shopping feature is limited, experimental, and U.S.-only. But the underlying idea is expansive: building AI agents that can execute user intent in the real world, starting with the most universal human activity of all, buying things.
Also read: Claude Sonnet 4.5 explained: Why Anthropic claims it’s the world’s best coding model
Vyom Ramani
A journalist with a soft spot for tech, games, and things that go beep. While waiting for a delayed metro or rebooting his brain, you’ll find him solving Rubik’s Cubes, bingeing F1, or hunting for the next great snack. View Full Profile