What is agentic commerce? A Shopify guide

Agentic commerce is shopping where an AI agent does the legwork for a person: it reads product pages, compares options and, with permission, buys. This guide explains what agentic commerce means for your Shopify store, which protocols carry it, and what an agent needs from your product pages.

What is agentic commerce?

Agentic commerce is shopping carried out by AI agents acting for a person. The agent takes a goal, such as a gift under a set budget, then searches, compares and can complete the purchase within limits the shopper set. The merchant still sells, ships and supports the order.

Merriam-Webster defines agentic, in its AI sense, as a system that “acts autonomously or with little human supervision.” McKinsey describes agentic commerce as shopping in which AI agents act on the shopper’s behalf. The Stripe guide describes AI agents “finding, comparing, and potentially making purchases.”

Here is how it differs from AI shopping you already know. An assistant that answers “which soap travels well?” is AI shopping. An agent that buys the soap, at the price the shopper approved, from the store it chose, is agentic commerce. Both start in the same place: what the agent can read about your products.

The stakes are large on paper. McKinsey estimates agentic commerce could orchestrate $900 billion to $1 trillion of US B2C retail revenue by 2030, and $3 trillion to $5 trillion worldwide, counting goods only. Treat that as a scenario for the market, not a forecast for your store.

How does agentic commerce work?

An agentic purchase runs in stages. The shopper states a goal and limits, the agent finds and compares products from merchant data, checkout runs through a protocol or the store’s own checkout, and payment moves as a token or signed mandate rather than the shopper’s raw card details.

The step-through below walks one purchase from the shopper’s request to your fulfillment, naming the published standard at each stage.

How an agentic purchase runs, step by step

1. 1 Goal

   The shopper tells the agent what they want and the limits: budget, timing, preferences. Google's Agent Payments Protocol records this as a signed Intent Mandate.

2. 2 Discover

   The agent reads structured product data. OpenAI's commerce docs describe ChatGPT ingesting structured catalog data from product feeds, and Shopify makes products available to AI channels through Shopify Catalog.

3. 3 Compare

   The agent weighs relevance, price, availability and what each product page says. When it launched shopping in ChatGPT, OpenAI said product results are organic and unsponsored.

4. 4 Checkout

   The agent starts checkout through a protocol such as ACP or UCP, or hands the shopper to your store's own checkout. Under both protocols you remain the merchant of record.

5. 5 Pay and fulfill

   Payment moves as a token, such as Stripe's Shared Payment Token or Mastercard Agentic Tokens, or under an AP2 Cart Mandate. Your store ships the order and handles support and returns as usual.

Two of these stages are yours to shape. The checkout stage depends mostly on your platform. The discover and compare stages depend on your data and your pages, and that is where an agent decides whether your product is the one it buys.

What is an example of agentic commerce?

The clearest examples come from the large platforms. Amazon announced a Buy for Me beta that buys from other brands’ sites, OpenAI tested and then scaled back in-chat checkout in ChatGPT, and Shopify’s agentic storefronts let eligible stores sell in ChatGPT, Google AI Mode, Gemini, Microsoft Copilot and Meta.

Each of these is described on the company’s own site:

• Amazon Buy for Me. On April 3, 2025, Amazon announced a beta for a subset of US customers in the Amazon Shopping app. A shopper taps Buy for Me on a product sold on another brand’s site, and Amazon uses agentic AI to buy it there with the shopper’s encrypted name, address and payment details. The brand sends the confirmation and handles delivery, returns and service.
• ChatGPT Instant Checkout. On September 29, 2025, OpenAI launched Instant Checkout for US ChatGPT users buying from US Etsy sellers, with Shopify merchants listed as coming soon. On March 24, 2026, OpenAI said the first version “did not offer the level of flexibility that we aspire to provide.” Merchants can now use their own checkout while ChatGPT focuses on product discovery, fed by merchants’ product feeds.
• Shopify agentic storefronts. Shopify’s help center says the feature is active by default for eligible stores. In ChatGPT, the shopper completes the purchase on your store’s checkout inside ChatGPT’s in-app browser, or in a new tab on ChatGPT on the web. In Google AI Mode, Gemini, Microsoft Copilot and Meta, with direct checkout activated, the shopper buys without leaving the channel.
• Perplexity Buy with Pro. At its November 2024 launch, Perplexity introduced one-click checkout for US Pro subscribers on select products from select merchants, with product data from integrations including Shopify.

What is the Agentic Commerce Protocol?

The Agentic Commerce Protocol, or ACP, is an open standard developed by Stripe and OpenAI under the Apache 2.0 license. It lets an AI agent start a checkout with a business and pass the buyer’s payment credentials without exposing them. ChatGPT was the first platform to implement it.

The ACP site calls it “an open standard for programmatic commerce flows between buyers, AI agents, and businesses.” Businesses keep control of which products can be sold, how they are presented and how orders are fulfilled. Stripe is the first compatible payment provider, through its Shared Payment Token. At launch, OpenAI said ACP keeps merchants “in control of the customer relationship as the merchant of record.”

ACP is one of several standards. Google published two of its own:

Protocol	From	What it covers	Your role
ACP	Stripe and OpenAI	Checkout and payment credentials between agents and businesses	Merchant of record
UCP	Google, with Shopify, Etsy, Wayfair, Target and Walmart	Checkout and product discovery, starting with AI Mode in Google Search and Gemini	Merchant of record
AP2	Google, with more than 60 organizations	Agent payments, proved by signed mandates	Gets a signed record of what the shopper approved

Google introduced the Universal Commerce Protocol on January 11, 2026, with endorsements from partners including Visa, Mastercard, Stripe, Adyen and American Express. Google’s merchant page says “You remain the Merchant of Record,” that native checkout is the default path, and that UCP works with AP2, A2A and MCP. Adoption runs through a waitlist.

The Agent Payments Protocol, announced September 16, 2025, proves what the shopper allowed. An Intent Mandate records the request and its limits. A Cart Mandate records the exact items and price the shopper approved.

On Shopify, agentic storefronts are how your products reach these channels: the help center says products are made available to AI channels through Shopify Catalog.

What is agent to agent commerce?

Agent to agent commerce is when a shopper’s agent transacts with another agent that acts for the seller, for example negotiating a bundle discount with a retailer’s agent. McKinsey names it as one of three interaction models, next to agent to site and brokered agent to site.

McKinsey’s report sets out the three models:

1. Agent to site. “Agents interact directly with merchant platforms,” such as an agent comparing hotel sites before it books.
2. Agent to agent. “Agents transact autonomously with other agents,” such as a shopping agent negotiating with a retailer’s agent.
3. Brokered agent to site. Intermediary systems coordinate several agents and platforms, such as a booking agent working through a platform’s broker agent.

The examples earlier in this guide work with a merchant’s site or a platform’s checkout. The payment standards are built with the agent to agent model in view: Google says AP2 can be used as an extension of the Agent2Agent (A2A) protocol and the Model Context Protocol (MCP).

What must merchants do so AI agents can read their product data?

Give agents product data they can read and trust: a complete catalog feed, a page that states what each product is and who it is for, answers to the questions shoppers ask, and crawler rules that let AI agents in. Then check what the assistants actually say about you.

Many retail pages fall short today. In April 2026, Adobe scored pages across US retail websites for how much of their content machines can read. Product pages scored lowest of the page types Adobe reported. Adobe does not publish how many pages it scored, so read the numbers as a benchmark, not a census.

Share of page content machines can read, by page type

Source: Adobe, U.S. retailers see surge in AI traffic (April 16, 2026)

Show data

Share of page content machines can read, by page type

Page type	Value (%)
Product pages	66
Store locator	73
Category pages	74
Homepages	75
FAQ pages	80
Returns pages	82

Five things are within your control:

1. Check your agentic storefronts setting. In your Shopify admin, open Sales channels, then Agentic, and see which AI channels your products reach.
2. Complete your product data. Stripe advises exposing “product, pricing, availability, and policy data in formats agents can query directly.” OpenAI’s feed guidance covers product copy and variant modeling.
3. Say it on the page. State what the product is, who it is for and what it is not, and answer the questions shoppers type. An assistant repeats what your page says, and guesses where it says nothing.
4. Let the agents in. Your robots.txt decides which AI crawlers may read your pages. The checker below shows the answer for your store.
5. Measure what the assistants say. Ask ChatGPT, Gemini and Perplexity the questions your shoppers ask, many times, because they do not give the same answer twice.

AI crawler access checker

Check whether 15 AI crawlers, from GPTBot to PerplexityBot, may read your site's pages.

Your website's address

For example yoursite.com.

Check my site

A quick check confirms you are a person, not a bot.

Descriptions checked 2026-10-13. We do not save the address or the result.

Open the full tool

One of our users, the Shopify skincare brand Jericho, shows why measuring comes first. When we first measured it in September 2026, ChatGPT, Gemini and Perplexity recommended it in 1 of 225 answers (0.44%). The product pages carried the merchant’s description and little else an assistant could use.

That loop is what we built SourceVision for. We ask the assistants your shoppers’ questions many times, propose the missing page content from facts you confirm, write nothing until you approve, and read every change back, with a record of what it replaced. See how it works, or try our free AI crawler checker on its own page.

SourceVision is in an invite-only beta. Request an invite.

What are the risks of agentic commerce?

The main risks are fraud, telling trusted agents from unwanted bots, and distance from the shopper. Agent purchases weaken the signals fraud tools rely on, bad automation can pose as an agent, and an agent can choose a product before a person ever sees your brand.

Fraud models change first. Stripe says agent payments need “customer modeling purpose-built for agent-initiated transactions, where traditional signals no longer apply,” and plans Radar updates to tell trusted agents from bots.

The card networks are building identity for agents. On October 14, 2025, Visa introduced the Trusted Agent Protocol, developed with Cloudflare, so merchants can tell approved agents from “malicious automation and rogue bots.” On April 29, 2025, Mastercard unveiled Agent Pay, which requires agents to be registered and verified and lets shoppers limit what an agent may buy.

Liability is less settled. The sources above do not say who carries a dispute on an agent purchase, so read your payment provider’s terms before you rely on any channel.

The relationship risk is the quiet one. Shopify says merchants keep “full ownership of the customer relationship and post-purchase experience” for orders from AI channels, and those orders show channel attribution in your admin. But the choice of product happens inside the agent. If your page leaves a question open, the agent fills the gap with a competitor’s answer, and you never see the shopper who was lost.