Shoppers are starting to ask AI assistants to find, compare and buy products for them. Here is what agentic commerce is, which standards are emerging, and a practical readiness plan for Canadian retailers and e-commerce brands.
For twenty-five years, e-commerce has followed the same pattern. A shopper searches, clicks through to a store, browses, adds to cart and checks out. Every step happens on the merchant’s website, under the merchant’s design and analytics.
Agentic commerce changes that pattern. A shopper tells an AI assistant what they want (“waterproof hiking boots, size 10, under $200, delivered by Friday”), and the agent finds options, compares them and, increasingly, completes the purchase without the shopper ever visiting the store’s product page.
This is no longer theoretical. OpenAI introduced its Agentic Commerce Protocol with Stripe in 2025 to power purchases inside ChatGPT. In January 2026 Google launched the Universal Commerce Protocol, co-developed with Shopify, Etsy, Wayfair, Target and Walmart, to let agents in Google’s AI Mode and the Gemini app check out on a shopper’s behalf, and expanded it in March 2026 with cart and catalogue capabilities.
This guide explains how agentic commerce works, what it means for retailers, and how to prepare, including in Canada, where rollout is still catching up.
Key takeaways
- Agentic commerce means AI agents discover, compare and buy products for shoppers, often without a visit to your website.
- Emerging standards include Google’s Universal Commerce Protocol (UCP) and OpenAI’s Agentic Commerce Protocol (ACP); merchants generally remain the merchant of record.
- Agents choose products using structured data: feeds, prices, availability, shipping, returns and reviews. Data quality becomes a ranking factor.
- Consistency between your site, structured data and product feeds is essential; mismatches can exclude you.
- Prepare now with clean feeds, complete policies, server-side order tracking and a plan to join programs as they open in your market.
What agentic commerce is
Agentic commerce is shopping in which an AI agent acts on a person’s behalf to discover products, compare options and complete purchases, using structured information and standard protocols provided by merchants and platforms.
It sits on a spectrum:
| Stage | What the AI does | What the shopper does |
|---|---|---|
| AI-assisted search | Suggests products and links | Clicks through and buys on your site |
| AI comparison | Builds a shortlist with prices, reviews and delivery dates | Chooses and checks out on your site |
| Agentic checkout | Completes the purchase inside the AI experience | Confirms the order |
| Delegated shopping | Buys within rules the shopper sets (reorders, price thresholds) | Sets preferences and budgets |
Most activity in 2026 sits in the first three stages. The fourth is emerging for low-risk, repeat purchases.
The emerging standards
Google’s Universal Commerce Protocol (UCP)
- Launched in January 2026, co-developed by Google with Shopify, Etsy, Wayfair, Target and Walmart, and endorsed by more than 20 partners including Visa, Mastercard, Stripe and Adyen.
- An open, vendor-agnostic standard for how agents and merchant systems communicate: checkout, identity linking for loyalty benefits, order management, and since March 2026, cart and product catalogue access.
- Merchants remain the merchant of record; UCP handles communication, not payment processing.
- It works alongside other standards such as the Model Context Protocol (MCP) and Agent2Agent (A2A).
- Initially open to eligible U.S. merchants, with checkout via Google Pay and global expansion planned through 2026.
OpenAI’s Agentic Commerce Protocol (ACP)
- Introduced with Stripe in 2025 to enable purchases directly inside ChatGPT.
- Lets merchants expose products and accept orders from ChatGPT while keeping their own payment processing and customer relationship.
For most merchants, the practical path into these programs runs through their commerce platform (such as Shopify or other major platforms) or payment provider, rather than building protocol integrations from scratch.
How agents decide what to buy
An AI agent cannot be charmed by a beautiful homepage. It decides based on information it can read and trust:
| Factor | What the agent looks at | What you control |
|---|---|---|
| Relevance | Product titles, descriptions, attributes (size, material, compatibility) | Complete, accurate product data |
| Price and value | Current price, promotions, total cost including shipping | Feed accuracy; transparent pricing |
| Availability | Stock levels by variant and location | Real-time or frequent inventory updates |
| Delivery | Shipping speed and cost to the shopper’s location | Clear shipping rules in feeds and on site |
| Policies | Returns, warranties, support contacts | Complete, consistent policy information |
| Reputation | Reviews and ratings, merchant trust signals | Review collection; service quality |
| Consistency | Whether site, schema and feed agree | Single source of product truth |
In agentic commerce, product data quality becomes a ranking factor. Two stores selling the same boots at the same price may be treated very differently if one has complete attributes, accurate stock and clear return terms and the other does not.
What changes for retailers
Fewer page views, more direct orders
If agents complete purchases inside AI experiences, some orders will arrive without a session on your website. Your analytics need to recognise and attribute these orders, usually through server-side order data rather than browser tracking. Google has noted that UCP transactions do not create site sessions or page views.
Brand experience shifts to data and service
Your product imagery and design still matter for shoppers who browse. But for agent-led purchases, your “brand experience” becomes the accuracy of your data, the speed of delivery, the ease of returns and the quality of support after the sale.
Competition on clarity
Agents compare options side by side. Hidden fees, vague delivery estimates and incomplete specifications are more likely to cost you the sale.
New partners in the funnel
Platforms, payment providers and AI companies are now part of the path to purchase. Expect new programs, fees and requirements to evolve quickly.
The agentic commerce readiness checklist
Product data
- Complete, accurate titles and descriptions written for clarity, not keyword stuffing.
- Full attributes for every variant: size, colour, material, dimensions, compatibility, GTIN where applicable.
- High-quality images with descriptive alt text.
- Prices and availability updated frequently, ideally in near real time.
Feeds and structured data
- A complete Google Merchant Center account and product feed, with no disapproved items left unresolved.
- Product, Offer and Review structured data on product pages that matches the feed exactly.
- Shipping and return policies configured in Merchant Center and visible on your site.
- For UCP participation where eligible, the attributes Google requires (such as a native-commerce flag per product) and a published capabilities manifest at
/.well-known/ucp.
Policies and service
- Clear, specific returns, exchanges and warranty terms.
- Delivery estimates by region, including for Canadian provinces and territories where you ship.
- Customer support contact details that work.
- Bilingual information if you sell in Quebec.
Operations
- Order management that can accept orders from multiple channels and keep inventory in sync.
- Fraud screening that recognises legitimate agent-initiated orders.
- Server-side tracking of orders by channel, so you can measure agentic sales.
Discoverability
- Strong SEO and Generative Engine Optimization for your brand and categories, so AI assistants know and trust your store before a shopper even asks.
- Reviews on your site and on third-party platforms.
- Consistent brand and business information across the web.
Product data: before and after
The difference between a product an agent can confidently recommend and one it skips often comes down to a few fields. Here is a simplified example for a pair of hiking boots.
| Field | Weak listing | Agent-ready listing |
|---|---|---|
| Title | “Trail Boot Pro” | “Trail Boot Pro Men’s Waterproof Hiking Boot, Brown, Wide Fit” |
| Description | “Our best boot yet. Perfect for adventure!” | Clear summary of use (day hikes, light backpacking), waterproof membrane, sole type, weight per boot, break-in notes |
| Sizes | “Multiple sizes” | Each size as a variant with its own stock status |
| Materials | Missing | Upper, lining, sole and insole materials listed |
| Price | Shown on site only | Same price in feed, structured data and page, with any sale dates |
| Shipping | “Fast shipping” | Delivery estimate by region, including Canadian provinces, and shipping cost |
| Returns | Link to a generic policy page | Specific return window, condition rules, who pays return shipping |
| Identifiers | Missing | GTIN and manufacturer part number |
| Reviews | A few on the page | Reviews marked up and syndicated where possible |
None of these changes require new technology. They require care and a single source of product truth that feeds your site, your structured data and your channels.
Payments, fraud and customer service in an agentic world
Agentic checkout introduces some new operational questions:
- Payment methods: early programs rely on specific wallets and providers. Make sure your payment stack can accept them as they become available in your market.
- Fraud rules: orders placed through agents can look unusual to traditional fraud filters (new device patterns, no browsing history). Work with your payment provider so legitimate agent-initiated orders are not declined.
- Customer communication: the shopper may never have seen your website. Order confirmation emails, delivery updates and packaging become the main brand touchpoints. Make them clear and reassuring.
- Returns and support: make it easy for customers to find who they bought from and how to get help, even if they ordered inside an AI assistant.
- Loyalty: identity-linking features in protocols such as UCP aim to let shoppers keep loyalty benefits when buying through agents. Plan how your loyalty program will work across channels.
A note for Canadian merchants
Many agentic checkout programs launched first in the United States. For Canadian merchants, that creates two practical priorities:
- Get the foundations right now. Clean product data, accurate feeds and clear policies help today in Google Shopping, AI Overviews and AI assistants, and they are prerequisites for checkout programs when they expand.
- Watch your platform’s roadmap. Shopify and other major commerce platforms are likely to be the main route for Canadian merchants into these programs. Follow their announcements and enable features as they become available in Canada.
Canadian-specific details matter in feeds and policies: prices in CAD, correct GST/HST and, where relevant, QST treatment, cross-border shipping and duties for U.S. customers, and French-language product data for Quebec.
A 60-day plan
| Weeks | Actions |
|---|---|
| 1 to 2 | Audit product data completeness; fix missing attributes on top-selling products; resolve Merchant Center errors |
| 3 to 4 | Align structured data, feeds and on-site prices, stock and policies; publish clear shipping and returns pages |
| 5 to 6 | Implement server-side order tracking by channel; review fraud rules for agent-initiated orders |
| 7 to 8 | Test how AI assistants describe and recommend your products; improve product content and reviews; register interest in agentic programs available to you |
How to measure agentic commerce readiness
- Feed health: percentage of products approved with complete attributes.
- Data consistency: mismatches between site, schema and feed (target: zero).
- AI visibility: how often AI assistants recommend your products for your key category prompts.
- Agent-originated orders: orders attributed to AI channels, once available.
- Post-purchase quality: delivery accuracy, return rates and review scores.
Frequently asked questions
UCP is an open standard launched by Google in January 2026 with major retailers and platforms. It lets AI agents in Google’s AI Mode and the Gemini app discover products, manage carts and check out with merchants, while merchants remain the merchant of record.
Availability depends on the program and platform. Early checkout programs focused on U.S. merchants, with expansion planned. Canadian merchants should prepare their data and follow their commerce platform’s announcements.
Protocols are designed to keep merchants as the merchant of record, so you fulfil the order and typically receive customer details needed for delivery and service. The browsing relationship on your site may be reduced, which makes post-purchase experience more important.
No. Many shoppers will continue to browse, research and buy on websites, especially for considered purchases. Your website also remains a key source that AI agents read. Agentic commerce adds a channel; it does not remove the need for a strong store.
Audit your product data and Merchant Center feed. Complete, accurate, consistent data is the foundation for every AI shopping experience, from AI Overviews to agentic checkout.
The bottom line
Agentic commerce rewards the merchants that make it easiest for software to understand and trust them: complete product data, honest prices, clear policies and reliable fulfilment. Those are good business practices anyway. The retailers that invest in them now will be ready when AI agents become a meaningful share of their customers.
Netxeno builds and optimises e-commerce stores for retail and e-commerce brands, from product data and feeds to performance marketing and AI visibility. Find out how ready your store is with a free audit.
