Agentic Commerce on Shopify: The 2026 Guide

The short answer: agentic commerce means an AI assistant shops for the customer. On Shopify in 2026 that happens in two places. Your products are syndicated by default to AI channels like ChatGPT, Microsoft Copilot, Google AI Mode and Gemini, and some of those channels now take the order inside the chat. Separately, Microsoft's Brand Agents can put an AI shopping assistant on your own storefront. Neither needs Shopify Plus. What decides whether any of it sells is the quality of the data underneath: the catalog, the structured data, the policies the agent quotes and the checkout it hands off to.

This guide is the map. It dates every change that matters, shows which company ships which surface and what you control on each, lays out the four-layer readiness stack, and links to the deeper guide for each part. Facts are checked against the vendors' own announcements and help pages, listed under Sources at the end.

What is agentic commerce for a Shopify store?

Agentic commerce

A purchase journey in which an AI agent performs discovery, evaluation and sometimes the transaction on a shopper's behalf: querying catalogs, comparing options against the shopper's stated constraints, and either completing checkout in the conversation or handing the shopper to the merchant's checkout.

If you want the platform-neutral version before the Shopify detail, what agentic commerce is and how it works walks through the four stages, the protocols and every surface live today.

The key word is agent. A product-finder quiz is not agentic: a human still clicks every step. An agent takes intent ("a fragrance-free moisturizer for sensitive skin under $40 that arrives by Friday") and returns a decision, often a single product with a buy button. The store's job shifts from persuading a browsing human to being legible to a system that reasons over structured data on that human's behalf.

For a Shopify merchant this plays out on two surfaces, and most confusion in 2026 comes from mixing them up. Off-site, AI channels such as ChatGPT, Copilot, AI Mode and Gemini read your product data and recommend you, or don't. On-site, an agent on your own storefront answers questions, recommends products and walks the shopper to checkout. Different companies ship the two surfaces, different settings control them, and they fail in different ways.

How this differs from running a normal online store, row by row (who browses, what the storefront is, checkout, attribution and returns), is set out in agentic commerce vs traditional e-commerce.

The vocabulary moves fast (UCP, ACP, MCP, direct checkout, agentic storefronts). If a term in this guide is new to you, the agentic commerce glossary defines each one in a sentence.

What changed in agentic commerce in 2025 and 2026?

Six dated events shaped what a Shopify store can do today. Brand Agents and Copilot Checkout are Microsoft's, announced in January, not part of Shopify's June edition. And buying inside ChatGPT, the most-quoted feature of 2025, no longer works the way it launched.

Read it as a shift in where the sale closes. In 2025 the bet was that the chat would become the checkout. By late 2026 it is mixed: Shopify's help center lists direct checkout inside Copilot, AI Mode, Gemini and Meta, while ChatGPT sends the shopper to the merchant. So the same catalog has to work for an agent that buys and an agent that refers. The Copilot Checkout configuration guide covers the channel where the checkout itself now runs inside the conversation. The AI shopping channels guide goes channel by channel, and if your products aren't showing up at all, start with every cause of products not showing in ChatGPT.

Which AI channels sell Shopify products, and what do I control?

Who ships what, where the sale closes, and what you actually control. Check it before you tell anyone "we're live in ChatGPT".

Two things catch people out. Every Shopify row sits behind one setting: in Sales channels → Agentic, "Allow Shopify to manage for me" is on by default, activating every channel and direct checkout and enrolling you in new ones. Turn it off to set Catalog access and direct checkout per channel. And the Brand Agents row has its own consent: accepting Microsoft's terms publishes your catalog to Microsoft Merchant Center, a second route into Copilot alongside Shopify Catalog. Keep the two in agreement.

How does selling in ChatGPT, Copilot and AI Mode work for a Shopify store?

Agentic storefronts

Shopify's name for the AI channels that receive a store's eligible product data through Shopify Catalog, managed under Sales channels → Agentic, where direct checkout can be turned on or off per channel.

Three Shopify pieces do the work. Shopify Catalog standardizes product data from eligible stores and makes it available to AI channels; per Shopify's Catalog requirements, a store must be on the Starter plan or higher and not in private mode, and each product needs a title, at least one image, a price above zero and an identifiable product URL. You can block individual channels, but you can't opt out of Catalog itself.

Universal Commerce Protocol (UCP)

An open standard, co-developed by Google with Shopify and major retailers, that gives AI agents a common way to search catalogs, build carts, run checkout and track orders with a business.

UCP is the language underneath. Shopify says its merchants are UCP-enabled by default, and its developer docs describe UCP-compliant MCP servers for catalog search, carts and checkout. Most agents create a checkout and hand the buyer to your storefront to pay; agents Shopify trusts can complete checkout directly. That is the mechanism behind Copilot's and Google's in-chat buying. Direct checkout is the merchant-facing result: a Shopify-powered checkout inside the AI conversation, on your payments, taxes and order flow, with orders landing in your admin.

ChatGPT is the exception. OpenAI launched Instant Checkout in September 2025 and, in a March 24, 2026 post, said the first version lacked the flexibility it wanted, so merchants now use their own checkout while OpenAI focuses on discovery. On Shopify that means ChatGPT shoppers finish on your online store, and there is no ChatGPT direct-checkout setting to change. Treat ChatGPT as a referral channel with very high intent, and make sure the page it lands on states the same price and stock the chat just quoted.

The checkout-side work, in Copilot especially, is about the unhappy paths: unsupported product types, delivery methods that don't exist in the chat, discount logic, tax. The Copilot Checkout configuration guide walks through each.

For the variants of in-chat and agent-completed checkout beyond Shopify, and the payment tokens behind them, see what agentic checkout is.

What are Microsoft Brand Agents, and how does a Shopify store get one?

Brand Agent

Microsoft's AI shopping assistant for a merchant's own website, set up and measured in Microsoft Clarity, that answers product questions in the brand's voice and guides shoppers from discovery to checkout.

Brand Agents is the on-site surface. Microsoft announced it on January 8, 2026 alongside Copilot Checkout, describing AI shopping assistants that speak in your brand's voice and guide customers from browsing to purchase on your own site. Microsoft's documentation lists it as a beta for Shopify and WooCommerce websites (not mobile apps), with an agent dashboard that compares agent-assisted sessions to organic ones.

Getting one on Shopify is a Microsoft process, not a Shopify setting. You install the Microsoft Clarity app from the Shopify App Store, turn on the Clarity script, choose Clarity with Brand agents and join the waitlist; Microsoft emails you when you're approved. Configuration covers the greeting, the look, and three pieces of brand identity Clarity pre-fills with AI-generated text: a brand overview, your offering, and tone. Edit those. Pre-filled copy about your own brand is exactly where a confident wrong answer starts.

Two consequences are easy to miss. Accepting the Brand Agents terms publishes your catalog to Microsoft Merchant Center, which makes products discoverable in Bing and Copilot. And Microsoft's terms require a privacy-policy disclosure that conversation content is shared with Microsoft; Microsoft publishes sample wording. Neither Microsoft nor Shopify lists a Shopify plan requirement. Microsoft Brand Agents on Shopify, explained is our canonical explainer: what the agent reads, what it can't do yet, and how to train it.

Is a Brand Agent just a chatbot?

No, and treating it like one is the most common way to waste it. A scripted chatbot follows a decision tree someone wrote: if the shopper says returns, show the returns macro. A Brand Agent generates answers from your catalog, site content and the brand identity you configure. That makes it far better at the questions no one scripted ("will this fit a 60-inch sofa and is it machine washable?") and far worse when the data behind it is thin, because it answers anyway.

Many stores will run both for a while: the helpdesk bot for order-status and returns workflows tied to support tooling, the Brand Agent for pre-purchase questions. If you're deciding whether to replace a Gorgias, Tidio or similar bot, the Brand Agents vs. chatbots migration guide covers what carries over, what breaks, and a sequence that avoids a support gap.

Do I need Shopify Plus for agentic commerce?

No. Nothing in the agentic stack is gated to Plus. Shopify Catalog requires the Starter plan or higher. The channel pages for ChatGPT, Copilot and Google name requirements about where you sell, Catalog eligibility, completed policies and (for Google) a Merchant Center account, not a plan tier. Microsoft names no Shopify plan for Brand Agents. And brands whose storefront isn't on Shopify at all can use the Agentic plan to sync products to Shopify Catalog.

The requirements that actually block stores are mundane. Policies not filled in under Settings → Policies. Products with no image or a zero price. Categories Shopify Catalog excludes, such as alcohol, tobacco, weapons and services. A store in development mode. Product descriptions where a legal disclosure sits past the first 6,000 characters, which Shopify's ChatGPT page says should be avoided. Check those before anything else; the agentic commerce readiness checklist has them as tick-boxes.

What does an agent-ready Shopify store need? The four-layer stack

Think of an agentic storefront as a stack. Each layer reads the one beneath it, so the weakest layer caps everything above. This holds for both surfaces: ChatGPT and Copilot read the same catalog your Brand Agent does.

Merchants get excited about layer 3 because the agent is the visible part. But an agent on a thin catalog is a confident liar. Build bottom-up. The readiness checklist turns each layer into checks you can run yourself in an afternoon.

Why is the catalog the real bottleneck?

Because every surface reads it and none of them can infer what isn't there. A human shopper forgives a blank material field; they look at the photo. An agent asked "is this 100% cotton?" either guesses or declines, and both lose the sale. Shopify makes the point itself: the Agentic page grades each product's listing quality on description completeness, image coverage, reviews, variant and option completeness and shop-policy completeness, and suggests fixes for products that didn't rank. In its Spring '26 announcement, Shopify also says AI searches powered by Shopify Catalog convert at twice the rate of searches on scraped data. That's Shopify's own figure, but the direction is the argument for structured data.

An AI-unready catalog usually has three problems: sparse attributes (half the fields a shopper would ask about are empty), prose specs (facts buried in paragraphs instead of fields), and no compatibility data ("fits", "works with", "suitable for"). The fix is covered in depth in the catalog enrichment playbook.

What happens when AI quotes the wrong price or delivery date?

It happens when the sources an agent reads disagree. Your product schema says one price, a Shopify automatic discount says another, a theme banner promises free two-day shipping that your carrier rules don't support, and the AI channel quotes whichever it read last. In a direct checkout the shopper sees the real total at the last step and leaves; through a ChatGPT referral they land on a page that contradicts the chat. Either way you lose the sale and some trust.

The fix is one source of truth per fact. Price and sale price come from the same field everywhere: product data, schema and any feed. Shipping cutoffs and delivery windows live in your shipping settings and policies, not in hard-coded theme text. Return windows are stated once, in Settings → Policies, which Shopify requires for AI channels anyway. Before peak season, compare what three AI channels say about five of your best sellers against your live product pages.

• Price drift: schema, discounts and marketplace coupons out of step. See when AI shopping quotes the wrong price.
• Delivery drift: stale cutoffs, banners and marketplace SLAs. See when AI promises a delivery date your warehouse can't hit.

How do I train an agent to answer policy and product questions?

Two tools shape what agents say about your store, and they come from different companies. Shopify's Knowledge Base app lets you preview the store facts Shopify generates, review the questions shoppers ask in AI conversations, and write your own answers. Per Shopify's help center, those FAQs don't show on your storefront; they're a data source AI platforms use, and they improve accuracy without changing how often you appear. Microsoft's Brand Agents take your brand overview, offering and tone in Clarity, on top of your catalog and site.

Start with the questions that cost money when answered wrong: returns and exchanges, shipping cutoffs and costs, sizing and fit, compatibility, care. Override the generated answers in plain, specific sentences ("30 days from delivery, unworn, free return label in the U.S."), and keep them identical to your policy pages. Training the Shopify Knowledge Base covers the app field by field.

For the on-site agent, design for conversation patterns rather than single answers: comparisons, constraints ("under $50, vegan, ships to Canada"), suitability ("for a beginner?"), objections and post-purchase questions. Brand Agent conversation design lays out 30 flows and the order to build them in.

How do I set up agentic commerce on Shopify, in order?

A good setup is sequenced, not parallel. In this order each step stands on a finished one:

1. 01

   Baseline your readiness

   Score catalog, structured data, product pages and checkout before you change anything, so you can show what improved. What the free Readiness Score grades explains the four dimensions.

2. 02

   Clear the eligibility blockers

   Policies completed, products with images and prices, no excluded categories, store out of development mode, disclosures early in descriptions.

3. 03

   Enrich the catalog

   Fill attribute gaps for your top sellers first, move prose specs into fields, add compatibility data. This is the longest task and the one with the most leverage.

4. 04

   Align structured data and offers

   One price, one delivery promise, one return policy, stated the same way in product data, schema, feeds and policy pages.

5. 05

   Decide channel by channel

   In Sales channels → Agentic, decide whether to keep "Allow Shopify to manage for me" or set Catalog access and direct checkout per channel. Test a real order in each direct checkout, including a discount and an unsupported product.

6. 06

   Configure the agents

   Override Knowledge Base answers; if you're in the Brand Agents beta, edit the pre-filled brand identity and add the privacy disclosure.

7. 07

   Instrument, then watch the conversations

   Set your baseline metrics before launch and read real conversations weekly for the first month.

How do I measure agentic commerce?

You now get first-party numbers from both surfaces. Shopify's Agentic page reports sales, orders, online-store sessions and online-store conversion for each AI channel, combining referral sales and direct-checkout sales, plus the search queries your products appeared for. Clarity's Brand Agents dashboard tracks engagement, conversions and sales impact, and compares agent-assisted sessions with organic traffic. Pair those with your analytics and track five numbers from day one:

• AI-channel sales and orders, per channel, from the Agentic page.
• Direct-checkout share: how much of that revenue closed in the chat versus on your store.
• Agent-assisted conversion vs. site average, from the Clarity dashboard. The headline number for the on-site agent.
• Attributed revenue, closed inside or downstream of agent conversations.
• Unanswered questions, from the Knowledge Base query log and agent transcripts: the to-do list for your catalog.

Attribution is the hard part: a shopper may discover you in Copilot, ask your Brand Agent a question, and buy a week later from an email. How to measure agentic commerce ROI sets out the five-metric stack, attribution rules and the CFO math.

What goes wrong with agentic commerce, and what should the first 30 days look like?

• Building the agent first. It's the exciting layer, so teams start there, on a thin catalog. Build bottom-up.
• Assuming "live in ChatGPT" means "selling in ChatGPT". Discovery depends on your data; the sale closes on your store.
• Leaving direct checkout untested. It's on by default in Copilot, AI Mode and Gemini. Place a real order with a discount code before a shopper does.
• Two catalogs telling two stories. Shopify Catalog and Microsoft Merchant Center both feed Copilot. Keep them in agreement.
• Launching without a baseline. Without before-and-after numbers you'll have opinions about ROI, not results.

Most problems surface as symptoms: wrong answers, off-brand tone, checkout failures. What breaks first: 10 failure modes of a bad agent install maps each symptom to its root cause and fix. After launch, the work is a weekly rhythm of reading conversations, fixing data and re-testing; the first 30 days after your agent goes live gives it week by week.

What should a Shopify merchant do this quarter?

Not everything. The highest-return move for most $100K–$10M stores is narrow: confirm what Shopify is already syndicating and where each channel closes the sale, get an honest readiness baseline, then fix the catalog for your top sellers. That alone moves you further than a rushed agent on weak data. Add the Brand Agent when the data underneath can support it.

A workable 90-day order: in the first two weeks, open Sales channels → Agentic, note which channels are active and where each closes the sale, and place a test order in every direct checkout. In weeks three to eight, enrich attributes and align price, delivery and return facts for the products that earn most of your revenue, and override the Knowledge Base answers shoppers ask most. In the last month, join the Brand Agents waitlist if you want the on-site surface, set your measurement baseline, and re-score. Then repeat the loop for the next tier of products.