Agentic Checkout by Platform: What Shopify, WooCommerce and BigCommerce Require

Most of the agentic-commerce coverage answers "how does an AI agent buy" in the abstract. The question I actually get from operators is narrower: what do I need on my platform for an agent to check out. This is the platform-by-platform answer, and it comes with a warning up front. Platform details here move fast, so I last verified all of this in August 2026, and some of it will have shifted by the time you read it.

One distinction runs through everything below. Failing a requirement does not make an agent abandon your cart mid-flow. It makes you ineligible, so you never enter the fast path at all. The agent falls back to driving your normal storefront checkout like a human, and that is where the real abandonment happens. Keep those two failure modes separate, because they have different fixes.

Agentic checkout requirements start with three gates

Before any platform specifics, every agent checkout gates on the same three things. Miss one and you are out before the transaction starts.

The feed gate. Most shopping agents read product data through Google Shopping, which is fed by Google Merchant Center. Two independent studies point the same way: an analysis published by Search Engine Land (run by Tom Wells of Peec AI, March 2026) found ChatGPT matched about 83% of its carousel products to Google Shopping's top organic results, and a separate Semrush experiment put the top ChatGPT product in Google Shopping's top three roughly 75% of the time. Both are third-party research, not Google's own numbers, but they land in the same place: if you are not in the feed, you are not in the answer. The full feed breakdown is in the product feed requirements piece.

The data and schema gate. Past the feed, the agent needs clean, structured product data to understand what you sell. BigCommerce puts this bluntly through its SVP of Product for AI, Sharon Gee, who has said that without structured product data you cannot answer the questions agentic systems need answered to surface your products. Product schema, review and rating markup, return-policy and shipping markup, and FAQ content are what make a page legible to an agent.

“

“But here’s the truth: without clean, structured product data, even the most advanced AI can’t deliver its promise.”

The payment-trust gate. An agent spending money on a shopper's behalf has to prove it was authorized. This is what Google's Agent Payments Protocol (AP2) exists for: it carries cryptographically signed mandates (an intent mandate, a cart mandate, a payment mandate) so a purchase is authorized and non-repudiable. Card networks and processors are building around this, and it is the layer that lets a merchant trust an agent's payment at all.

“

“Merchants receiving agent payments must verify three things: (1) the payment was authorized by a real human (or a valid pre-authorization), (2) the payment settled successfully, and (3) the requesting agent is who it claims to be.”

The protocols behind agentic checkout: MCP, UCP, ACP and AP2

You do not need to implement these yourself; your platform does. But the acronyms are all over the release notes, so here is the clean mental model. They stack, they do not compete.

• MCP (Model Context Protocol), from Anthropic, is the data and read layer. An MCP server exposes tools like search products, get inventory, build cart, so any compliant agent can read your catalog and act on it.
• UCP (Universal Commerce Protocol), a joint Google and Shopify effort announced in January 2026, is the full-journey orchestration layer: discovery, checkout, and post-purchase handoff.
• ACP (Agentic Commerce Protocol), built by OpenAI with Stripe and announced in September 2025, handles checkout, the product feed, and payment delegation for the ChatGPT lineage.
• AP2 (Agent Payments Protocol), from Google with a large partner list, is the payment-authorization layer described above.

One correction worth making, because it changed the landscape: OpenAI did not kill in-chat checkout. In March 2026 it pulled back native Instant Checkout and pivoted toward product discovery plus merchant-owned checkout, while expanding ACP. Purchases from ChatGPT now generally complete on the merchant's own store. That single move is why the per-platform picture below looks the way it does.

What Shopify requires for agentic checkout

Shopify is the most prescriptive, and it publishes an actual checklist. Its Agentic Storefronts launched in January 2026 and syndicate products to ChatGPT, Microsoft Copilot, Google AI Mode and Gemini from one admin.

To be eligible for direct (built-in) checkout, Shopify's Help Center lists three things: your products must be eligible for the Shopify Catalog, you must agree to the Agentic Storefronts supplemental terms, and you must have three policies completed in Settings then Policies - Terms of service, Privacy policy, and Return and refund policy. Miss the policies and you are not eligible for direct checkout.

Catalog eligibility is its own list: each product needs a title and at least one image, a price above zero, must ship to the United States or Canada, and must be published (not hidden, not password-protected). On the Agentic plan, where there is no Shopify online store, products must include an external product URL.

Here is the nuance most write-ups get wrong. As of now, ChatGPT on Shopify is discovery-only and always redirects the buyer to your online-store checkout, regardless of your policies. The channels that actually offer direct, built-in checkout are Microsoft Copilot, Google AI Mode and Gemini. So do not read "set your policies or ChatGPT sends buyers away" - ChatGPT sends them to your own checkout either way. The three-policy gate is about direct checkout on the Copilot and Google surfaces, and direct checkout currently displays only to United States customers.

What WooCommerce requires for agentic checkout

WooCommerce is more do-it-yourself, which fits its audience. The one thing it does natively is MCP: WooCommerce 10.3, announced in October 2025, shipped an MCP integration as a beta that is disabled by default. You enable it under WooCommerce, Settings, Advanced, Features, and it authenticates with WooCommerce REST API keys. That gives agents a native read-and-act surface on your catalog and orders.

The rest is add-ons. WooCommerce was not part of the UCP launch and has no core ACP support. It reaches ACP two ways: through third-party checkout plugins, and as a platform partner in Stripe's Agentic Commerce Suite, which handles the checkout and payment tokenization so the agent never touches card data. UCP support likewise comes from a third-party plugin plus a Google Merchant Center feed. Anyone telling you WooCommerce "supports ACP" natively is conflating the Stripe partnership with core support - worth knowing before you scope the work.

So the WooCommerce requirement is really: turn on MCP if you want agents reading your store, add a Stripe or plugin path for actual checkout, and keep a clean Merchant Center feed for the Google surfaces. There is no published Woo rule mandating specific schema, but the feed and data gates above still apply.

What BigCommerce requires for agentic checkout

BigCommerce (whose parent recently rebranded to Commerce) has gone broad on protocols. It ships a Storefront MCP server that exposes its catalog to AI agents. It endorsed Google's UCP in January 2026, planning buyable listings in AI Mode and Gemini for eligible US retailers. And in December 2025 it integrated Stripe's Agentic Commerce Suite so a single connection exposes your catalog to ChatGPT, Gemini, Copilot and others, with the merchant staying merchant of record.

On the data side, BigCommerce leans on Feedonomics, whose Agentic Catalog Exports push an optimized catalog to the agentic surfaces. Its own guidance is the clearest statement of the data gate of any platform: mark up your pages with Product, Review, AggregateRating, MerchantReturnPolicy and ShippingDeliveryTime schema, add FAQ content, and keep the data complete and current, or agents cannot reliably surface your products.

The BigCommerce requirement, then, is less about flipping one setting and more about catalog hygiene plus picking a path (Storefront MCP, the Stripe suite, or the UCP route) for the actual transaction.

Adobe Commerce and Salesforce agentic checkout support

Two more worth a line, because enterprise operators ask. Adobe Commerce is the most protocol-complete of the group: in February 2026 it committed to UCP and ACP on top of an earlier AP2 commitment, and it ships a Commerce MCP server. It frames this as a platform capability rather than a merchant checklist, so there is no Shopify-style list of prerequisites to point at.

Salesforce announced its largest Agentforce Commerce release around July 2026, taking Shopper, Buyer and Merchant agents to general availability with ChatGPT checkout via ACP live and Google plus Gemini described as coming. Requirement-wise it comes down to being on Salesforce Commerce with your catalog synced. Salesforce also markets that AI influenced 20% of global online sales in the 2025 holiday season, worth about 262 billion dollars, and that retailers running their own shopper agents grew sales 59% faster - both are Salesforce's own figures, so treat them as vendor data, not independent benchmarks.

When you fail a gate, the agent drives your checkout

This is the part that ties agentic checkout requirements back to lost revenue. Failing a gate does not throw an error the shopper sees. It quietly drops you out of the fast, protocol-based path, and the buyer is sent to your ordinary storefront checkout instead. That checkout then gets completed the old way - by the shopper in an in-app browser, or by a computer-use agent driving the page like a person. Either way you are now on the browser-driving path, and that is where the real abandonment happens.

That fallback path is exactly where browser-driving agents break on real stores: JavaScript-gated add-to-cart buttons, CAPTCHAs, multi-step forms an agent cannot fill. The two checkout architectures and how each one fails are in the checkout mechanics piece, and the missing-attribute failures that strand a sale are in the cart abandonment piece. The requirement gets you into the fast lane; the fallback is where you actually lose the order.

How to get your platform ready for agentic checkout

1. Clear the feed gate first. Get every product into Google Merchant Center with the required attributes (id, title, description, link, image_link, availability, price). Add product identifiers where they apply, and make sure a clear return policy is present on your site, since Google enforces that under its misrepresentation policy. Details in the feed piece.
2. Do your platform's specific setup. Shopify: complete the three policies and confirm Catalog eligibility. WooCommerce: enable the MCP beta and add a Stripe or plugin checkout path. BigCommerce: fix your schema and pick a transaction path. Salesforce and Adobe: sync the catalog and lean on the platform's protocol support.
3. Mark up your product pages. Product, Review, AggregateRating, MerchantReturnPolicy and ShippingDeliveryTime schema plus FAQ content. This is the data gate, and it is the same work whichever platform you are on.
4. Test the fallback. Assume some agents will drive your normal checkout. Make sure add-to-cart and checkout work without a human clicking through JavaScript gates. That is your abandonment insurance.
5. Measure what agents actually buy. Agent purchases hide from standard analytics, so instrument for them - the how is in the agent-revenue piece.

The pattern across every platform is the same underneath the acronyms: a clean feed, structured product data, a supported checkout path, and a storefront that survives an agent driving it by hand. Platforms differ on how much they do for you, but none of them clears those gates on your behalf. The whole cluster ladders up to the pillar guide.

Want to see whether an AI agent can already find and buy your products today? Run GEOlikeaPro's Visibility Vitals checker and start with the feed.