Agentic Commerce
Commerce where software agents can research, compare, prepare carts, and—in supported, authorized flows—help transact under shopper or merchant instructions.
What is agentic commerce?
Agentic commerce is commerce in which software agents can research and compare products, interact with catalogs, carts, and checkouts, and—in supported implementations—transact under shopper or merchant instructions. The term describes a range of delegated actions, not universal autonomy; confirmation and merchant controls remain essential.
An agent may retrieve products, compare sources, prepare a cart, or hand off to checkout. Transaction capability still depends on identity, payment, consent, policy, and jurisdiction. Use Shopify’s Agentic Storefront overview for merchant channels and the ACP explainer for protocol context.
How does agentic commerce differ from ordinary ecommerce?
Traditional ecommerce leaves navigation and checkout to the shopper; conversational shopping adds dialogue; agent-assisted research performs multi-step comparison; agentic checkout can take authorized transaction actions. These modes can coexist in one journey, and calling all of them “autonomous shopping” hides meaningful differences in risk and control.
| Mode | What software does | Transaction path | Human and merchant control |
|---|---|---|---|
| Traditional ecommerce | Displays pages, search, filters, and checkout | Shopper navigates and submits | Shopper chooses every step |
| Conversational shopping | Answers questions and surfaces options | Usually hands off to a store | Shopper reviews suggestions |
| Agent-assisted research | Collects sources, compares products, may prepare a cart | Handoff or cart creation | Buyer confirms product and merchant |
| Agentic checkout | Executes supported cart or payment actions | Integrated, authorized checkout | Explicit consent, policy, and transaction controls |
The dividing line is whether tools can change commerce state, under whose instructions and confirmation. See AI shopping agents for the journey-level view.
How does ChatGPT work with Shopify now?
ChatGPT shopping now uses Agentic Commerce Protocol discovery data, and Shopify Catalog is already integrated. Individual Shopify merchants generally need no extra discovery work; eligible shoppers currently finish through the merchant’s full Shopify checkout in the in-app browser or a new tab, not universal in-ChatGPT checkout.
OpenAI’s March 24, 2026 update explains ACP’s expanded discovery role. Shopping research can combine merchant data, public product information, and retail sources, so product results remain contextual.
For checkout, OpenAI’s Shopify merchant guidance points buyers to the store’s full Shopify checkout. Read how to sell on ChatGPT without repeating the old claim that every Shopify transaction completes inside ChatGPT.
Which historical launches demonstrated agentic checkout?
The September 2025 Etsy Instant Checkout launch and December 2025 Instacart partnership are specific examples of ACP-supported transaction experiences. They demonstrate that agentic checkout can exist, but they do not establish identical behavior for every platform, merchant, product, geography, payment method, or Shopify store.
OpenAI’s historical Instant Checkout announcement named Etsy availability and described Shopify as “coming soon.” The later Instacart partnership documented an integrated grocery experience. Use both as dated implementations, not current universal Shopify setup instructions.
The ChatGPT Instant Checkout guide keeps those launches separate from today’s merchant-owned Shopify checkout.
What can custom Shopify commerce agents do?
Shopify’s developer stack supports Universal Commerce Protocol, Model Context Protocol, catalog tools, and cart and checkout capabilities for custom agents. Eligible implementations can support deeper discovery or direct completion, but this is engineering capability requiring permissions and controls—not default autonomous checkout enabled for every merchant.
Start with Shopify’s agent developer documentation and carts and checkout APIs. Define what the agent may read, create, update, and submit; which identity it represents; and where buyer confirmation occurs.
A cart API call is not permission to substitute a product, accept new terms, or complete payment silently. State-changing actions need current price, inventory, variant, merchant, policy, consent, fraud, and error checks.
What product data and policies does agentic commerce require?
Agentic commerce requires accurate product identity, options, images, price, inventory, merchant information, and policies at the moment of action. Structured data, reviews, crawler access, or llms.txt can support readiness, but none makes an agent select a product or guarantees discovery, checkout, ranking, citation, or sales.
Shopify’s product eligibility guidance and OpenAI’s commerce policies define participation constraints. Shopify’s privacy guidance should inform data handling.
Merchants should maintain:
- Stable product and variant identifiers.
- Fresh regional price, availability, and delivery terms.
- Complete Terms, Privacy, Refund, shipping, and product disclosures.
- Authentic claims, comparisons, and reviews with material context.
- Catalog mapping for custom product fields or grouping.
Use the structured-data guide for consistency, not selection promises. Google’s helpful-content guidance likewise favors reliable, people-first information.
Which safeguards should a merchant require?
Require explicit approval for consequential actions, least-privilege access, fresh inventory and price checks, secure credentials, prompt-injection defenses, fraud controls, privacy limits, audit logs, idempotency, error handling, and rollback. Controls should become stricter as an agent moves from reading information to changing carts, payments, orders, or accounts.
- Define allowed read and write actions for each agent and channel.
- Revalidate product, variant, merchant, price, inventory, and terms.
- Require buyer confirmation before cart substitution or payment.
- Isolate untrusted content from system and tool instructions.
- Apply identity, fraud, rate, and payment controls.
- Minimize personal data and restrict order access.
- Log sources, calls, state changes, approvals, and outcomes.
- Prevent duplicate orders and test partial failure, cancellation, and rollback.
The NIST AI Risk Management Framework offers a governance reference. Map it to the deployed architecture and applicable obligations; citing a framework is not evidence that a live system is secure.
How should merchants measure agentic commerce?
Measure each state separately and carefully: eligible catalog record, retrieval, comparison display, cart creation, checkout handoff, payment authorization, completed order, cancellation, return, and support outcome. Preserve channel, product, variant, merchant, price, catalog version, consent, and timestamp without collecting unnecessary personal information.
Attribution remains conditional: referral plus order does not prove causation, and appearance does not prove attention or selection.
Use the AI crawler checker for open-web access only. Crawlability does not prove Catalog inclusion, agent selection, cart reliability, secure checkout, citations, or revenue.
What can StoreCited verify?
Run the free StoreCited readiness scan for a visible, point-in-time Shopify page and AI-search readiness audit. StoreCited does not see live agents or current product selections, control Catalog status, execute carts or payments, verify deployed security, or guarantee inclusion, ranking, citation, traffic, and sales.
Use the StoreCited overview to understand that boundary. Readiness evidence can identify storefront work, while platforms, merchants, developers, payment systems, policies, and buyers govern live agentic commerce.