Definition

Agentic commerce Agentic commerce is shopping in which an AI agent acts for a person: it finds and compares products and can complete the purchase, sometimes without the shopper opening the merchant's website. Open protocols such as the Agentic Commerce Protocol from OpenAI and Stripe and Google's Universal Commerce Protocol give agents and merchants a shared way to handle the order.

How agentic commerce works

In traditional ecommerce the shopper does every step: search, open product pages, compare, fill a cart, and check out. In agentic commerce an AI agent does some or all of that from a single request, such as "find me waterproof trail running shoes under $150 in a size 10 and order the best-reviewed pair."

A full agentic purchase moves through four stages:

1. Intent. The shopper describes what they need in plain language, with constraints like a budget or a delivery date.
2. Discovery. The agent searches merchant feeds and the web, often running several searches at once. This research step is agentic search.
3. Comparison. It weighs price, reviews, specs, and stock, then recommends one product or a short list.
4. Checkout. With the shopper's approval, the agent places the order with the merchant using saved payment details.

Most AI shopping features today stop after comparison and hand the shopper to the merchant's site. Checkout inside the conversation is harder to run, and it's the stage the commerce protocols were built for.

Agentic commerce protocols

For an agent to buy from a merchant, both sides need a shared way to exchange the order and the payment credentials. Two open standards lead that work.

Under both protocols the merchant stays the merchant of record and keeps its own checkout logic and customer data. Our guide to the Universal Commerce Protocol explains how UCP works and what adopting it takes.

The Model Context Protocol is a related but broader standard. It connects AI assistants to outside tools and data of any kind, while the commerce protocols cover the purchase itself.

Agentic commerce examples

• ChatGPT. OpenAI launched Instant Checkout in September 2025 so US users could buy from Etsy sellers without leaving the chat. In March 2026 it changed course: shoppers still find products in ChatGPT, but they usually finish the purchase on the merchant's own store, which opens inside the app ( Digital Commerce 360 covered the Shopify side of the change).
• Google AI Mode and Gemini. Through UCP, shoppers in the US can buy from participating retailers inside the conversation. Google opened it to a limited set of merchants first, with Etsy and Wayfair among the earliest.
• Amazon Buy for Me. Amazon began testing this feature in its Shopping app in April 2025. It buys products from other brands' websites for the shopper. We look at what it signals in Amazon's Buy for Me and the transition to agentic commerce.

The ChatGPT reversal says a lot about where things stand. Getting merchants onboarded and product data accurate turned out to be harder than building the chat, so for now the AI engines are strongest at discovery.

Agentic commerce vs AI shopping assistants

The two terms overlap, and the difference is whether the AI acts. An AI shopping assistant recommends: it answers "which blender should I buy" and links to product pages, the way ChatGPT Shopping product cards do. An agentic commerce system goes on to add the item to a cart and complete the order once the shopper approves.

Both sit on the same path. The recommendation step decides which products the agent even considers, so a brand that isn't recommended never gets to checkout.

Why agentic commerce matters for brands

When an agent does the comparing, the decision moves off your product page. The agent reads your data and weighs it against competitors before it shows the shopper a short list, so accurate, machine-readable product facts count for more than on-site design.

• Product data quality. Agents rely on feeds and structured data for price, availability, sizing, and specs. A missing or stale field can drop a product out of consideration.
• Protocol readiness. If your commerce platform supports ACP or UCP, an agent can complete checkout with you directly. Ask your platform before you build anything yourself. If you sell through a marketplace, the marketplace makes that choice for you, so check whether it has joined either protocol.
• Being recommended. Agents build their short lists from the same sources AI engines use for shopping answers, such as reviews and comparison articles.

How to track your brand in agentic commerce

You can't see inside an agent's decision, but you can track the recommendation step that feeds it. Run the shopping prompts your buyers would give an agent and record which products and merchants appear, how the engine describes them, and which sources it relied on.

Searchable's shopping analytics records the product cards and shopping results AI engines return for buyer prompts, and compares your product and merchant share with competitors. For a quick look at a single store, the free ChatGPT product visibility report shows how ChatGPT presents your products today.

Frequently asked questions

What is the difference between agentic AI and agentic commerce?
Agentic AI is any AI system that takes actions toward a goal instead of only answering questions. Agentic commerce is agentic AI applied to shopping, where the actions include filling a cart and paying.

What is the Agentic Commerce Protocol?
The Agentic Commerce Protocol (ACP) is an open standard that OpenAI and Stripe released on September 29, 2025. It defines how an AI agent passes an order and a scoped payment token to a merchant. It first powered Instant Checkout in ChatGPT, which OpenAI scaled back in March 2026.

What companies are using agentic commerce?
Google offers checkout inside AI Mode and the Gemini app through its Universal Commerce Protocol, starting with a limited group of US merchants that includes Etsy and Wayfair. OpenAI tried in-chat checkout in ChatGPT and now sends most purchases to the merchant's own store. Amazon has tested Buy for Me, which buys from other brands' sites inside the Amazon app.

Does the merchant still own the customer in agentic commerce?
Yes, under both ACP and UCP. The merchant stays the merchant of record and keeps control of pricing and the customer relationship, while the agent handles the conversation and passes the order along.

How can brands prepare for agentic commerce?
Start with product data, because agents read feeds and structured data for price and stock. Then ask your commerce platform whether it supports ACP or UCP, and track whether AI engines recommend your products for the shopping prompts your buyers use.

What are the risks of agentic commerce?
The main risks are an agent buying the wrong item, payment details being exposed, and decisions made on stale product data. Checkout flows built on ACP and UCP ask the shopper to confirm the order and pass a scoped payment token instead of full card details, and the merchant stays responsible for the sale. Stale data is the merchant's to fix, because an agent is only as accurate as the feed it reads.

Related terms

• Universal Commerce Protocol (UCP) An open standard backed by Google that lets AI agents discover products and complete purchases with merchants through a common interface.

• ChatGPT Shopping Product results inside ChatGPT answers, shown as cards with images, prices, and links to merchants.

• Model Context Protocol (MCP) The Model Context Protocol (MCP) is an open standard for connecting AI applications to outside tools and data. An MCP server describes what it can do in a format any MCP-compatible assistant understands, so one integration can work across AI apps like Claude and Cursor without custom code for each.

• Structured data Code, usually schema.org JSON-LD, that labels what a page contains, such as a product or an FAQ, in a format machines can parse.

• AI visibility How often AI engines bring up your brand when people ask about your category, and how accurately they describe it.

• Answer engine A search tool that replies with a written answer instead of a list of links. ChatGPT, Perplexity, Gemini, and Google AI Overviews are all answer engines.