Skip to main content
Agentic commerce — also called AI shopping — is shopping carried out by an AI agent on a customer’s behalf. Instead of browsing a storefront, the shopper asks an assistant like ChatGPT, Gemini, Perplexity, or Claude to find, compare, and buy a product. The agent reads live catalog data, narrows the options as the shopper refines what they want, and completes the purchase — sometimes on your store’s checkout, sometimes without the shopper ever leaving the chat. For a merchant, the question shifts from “how does my store rank in search” to “can an agent read my catalog and act on it correctly.” This page explains how a shopper uses an AI agent to buy, the protocols that make it work, and what you can do today to prepare.

How a shopper uses an AI agent

From the shopper’s side, buying through an agent is a conversation, not a sequence of pages. A typical purchase moves through three stages.
1

Describe what they want

The shopper states a need in plain language — “a waterproof jacket for autumn hiking under 1,500 kr” — instead of typing keywords into a search box. The agent reads structured product data from connected stores and returns a short, relevant set of options rather than a long results page.
2

Compare and refine the criteria

The shopper reacts and adjusts: “only breathable ones,” “something in green,” “what can ship to Denmark this week.” The agent re-queries live data — price, variants, availability — and updates its suggestions. This back-and-forth replaces filtering and tab-switching, and it’s where most of the decision happens.
3

Place the order

Once the shopper settles on a product, the agent guides them through checkout. Depending on the AI client and what your platform supports, this happens in one of the two ways below.

Where checkout happens

The final step — paying — takes one of two forms.

Embedded web checkout

The agent opens your store’s real checkout inside an in-app browser. The shopper completes the purchase on your page, using your payment, tax, and shipping setup. This is how ChatGPT handles checkout today.

Native agent checkout

The shopper confirms and pays without leaving the chat. The agent passes the order to your store through a commerce protocol, and your platform handles payment, tax, fraud checks, and fulfilment behind the scenes. Microsoft Copilot and AI Mode in Google Search work this way.
Either way, the order lands in your normal admin and you stay merchant of record — you keep the customer relationship, the payment, and post-purchase control. The only difference is where the shopper taps buy.

What changes for merchants

Discovery moves into the chat

Shoppers increasingly start with an assistant rather than a search engine. Your products are surfaced by an agent reading structured data, not by a page a person scrolls.

Machine-readable beats pretty

An agent does not see your design. It needs complete, consistent, structured product data — attributes, categories, pricing, and availability it can parse.

The customer relationship

Platforms differ on who owns the buyer. Shopify, for example, keeps the merchant as merchant of record, so you retain the customer relationship and post-purchase control.

Attribution and analytics

Orders placed through an agent are attributed to the AI channel, giving you insight into which assistants drive sales and what customers ask for.

The protocols

Several standards are emerging, each connecting AI agents to a store’s live data and checkout. They are young and moving quickly, and in most cases your e-commerce platform builds the integration so you don’t wire it up yourself.
ProtocolBacked byWhat it does
Model Context Protocol (MCP)Open standardBridges an agent to live store data such as inventory, pricing, and availability.
Universal Commerce Protocol (UCP)Shopify and GoogleDefines how agents create carts, check out, pay, and handle post-purchase. Powers native checkout in Microsoft Copilot, Gemini, and AI Mode in Search.
Agentic Commerce Protocol (ACP)OpenAI and StripeConnects ChatGPT to merchant catalogs for product discovery and purchase inside a chat.
No single protocol has won. The practical takeaway is to keep your product data clean and let your platform handle the protocol integrations as they ship.

How to prepare

The single most impactful step is to make your product data as clean, complete, and structured as possible. An agent can only sell what it can reliably read.
  • Complete attributes. Fill in size, color, material, and identifiers like GTIN or barcode for every variant.
  • Consistent categories. Group products so an agent can infer what they are and when to suggest them.
  • Accurate pricing and availability. Keep stock and prices in sync so an agent never quotes something you can’t fulfill.
  • Structured data. Add product schema to your pages so the underlying data is machine-readable.
  • HS codes for international sales. Correct customs tariff codes keep cross-border orders shippable. See understanding customs.
Most of this is the same data hygiene that improves traditional search ranking. Preparing for agents rarely means new work — it means finishing the catalog you already have.

On your platform

Both platforms Smart Send supports have published their direction for AI-driven shopping. Apply the data hygiene above, then dig into the store-side setup with our Shopify recommendations and WooCommerce integration guide.

Shopify Agentic Storefronts

How Shopify syndicates your catalog across AI agents from a single setup.
https://mintcdn.com/smartsend/bqSp1rejr9w3k8P_/icons/woocommerce.svg?fit=max&auto=format&n=bqSp1rejr9w3k8P_&q=85&s=45100db367d50450edd6e4aaee26fc1a

WooCommerce AI product discovery

How WooCommerce is building toward MCP, ACP, and UCP to keep stores discoverable.