Agentic commerce means AI agents that search, compare and buy products on a shopper’s behalf, and the Universal Commerce Protocol (UCP) is an open standard announced by Google to let those agents talk to merchants in a common way. For online stores, preparing for it is mostly about clean product data, machine-readable policies and a checkout that agents can complete. The work overlaps heavily with good feed management and technical SEO.
What is agentic commerce?
In traditional e-commerce, a person searches, clicks, reads and checks out. In agentic commerce, a person gives an instruction, such as ‘find running shoes under 120 euros that ship to Istanbul by Friday’, and an AI agent does the searching, comparing and sometimes the purchase. ChatGPT, Google’s AI Mode and Gemini, Perplexity and others have all shown shopping features along these lines.
The agent needs reliable answers to simple questions: what is the product, what does it cost, is it in stock, how fast can it ship, what is the return policy and how do I pay. Stores that answer those questions in a structured form are easier for agents to choose.
What is UCP, and how does it relate to other protocols?
Google introduced the Universal Commerce Protocol as an open standard for agent-to-merchant interaction, covering discovery, checkout and post-purchase steps, developed with retail and payments partners. It sits alongside other initiatives, such as OpenAI and Stripe’s Agentic Commerce Protocol used for checkout in ChatGPT, and payment-side efforts for agent authorisation.
These standards are young and still changing. Details of which platforms support which protocol, in which countries, shift month by month, so check current documentation before planning. The practical point for a merchant is that most of the groundwork is the same regardless of protocol: accurate product data, clear policies and a commerce platform that keeps up with integrations.
Why should a store care now?
Agents choose from what they can read. If a competitor’s feed lists stock, shipping times and return windows clearly and yours does not, the agent will tend to recommend the competitor because it can answer the shopper’s question with confidence. This is similar to how Google Shopping rewards complete feeds today.
There is also a measurement shift. Some purchases will happen without a visit to your site, which changes how you attribute sales and how you think about landing pages.
How Alien Road prepares a store for agentic commerce
Our process follows the same four stages we use elsewhere: discovery, setup, measurement and reporting. The difference is the reader. We are writing for software agents as well as people.
Discovery
- Audit the Google Merchant Center feed: titles, attributes, GTINs, availability, shipping and return settings.
- Check structured data on product pages: Product, Offer, shipping details and return policy markup.
- Review the commerce platform, such as Shopify, WooCommerce or Shopware, and which agent or protocol integrations it supports.
- Test the brand’s products in AI shopping surfaces with realistic prompts and record what the agents say.
Setup
We fix feed gaps first, since the same feed powers Shopping ads, free listings and, increasingly, AI shopping features. Product titles get the attributes shoppers ask about, such as size, material and compatibility. Shipping and return policies are set at account level in Merchant Center and marked up on the site.
Where the platform offers an agent checkout integration, we help the client evaluate it: which payment methods it supports, how orders flow into the order management system and what happens with customer data and consent. A typical engagement launches with a limited product range before expanding.
Measurement
Agent-driven orders need their own tracking. We tag them by source in the order system where integrations allow, monitor referral traffic from AI assistants in analytics and keep a monthly prompt test that checks whether the brand appears, with what price and with what claims.
Reporting
Monthly reports cover feed health, agent visibility in the prompt tests, orders and revenue from agent channels and any incorrect information agents showed about the client’s products. Errors usually trace back to a data field, which becomes the next fix.
An example scenario
Consider a WooCommerce store selling outdoor gear. Its prompt test shows an assistant recommending its tent but quoting last season’s price and saying returns are not accepted. The cause is a stale price in a secondary feed and a missing return policy in Merchant Center. Both are fixed in an afternoon. The next month’s test shows the correct price and a 30-day return window, matching the store’s real policy. Small fixes like this are most of the work, and they compound across hundreds of products.
What are the risks and open questions?
- Margin pressure: agents compare on price and delivery, which can favour discounting.
- Brand loss: the shopper may never see your site, design or story.
- Data ownership: who holds the customer relationship when an agent places the order.
- Errors: agents can state wrong prices or stock if feeds lag behind the site.
- Standard churn: early protocols may change or merge, so avoid custom builds that lock you in.
None of these is a reason to wait, but each is a reason to start with a limited pilot and keep your own site and customer channels strong.
What should a checklist for the next quarter include?
- Zero critical errors in Merchant Center diagnostics.
- Shipping and returns configured in Merchant Center and marked up on product pages.
- Product titles and descriptions that answer the top five questions customers ask.
- A monthly prompt test across at least two AI shopping surfaces.
- A decision on whether to join any agent checkout programme your platform supports.
Where to start this week
Ask two or three AI assistants to recommend a product you sell, using the words a customer would use. Write down whether your brand appears, what price is shown and whether the shipping and return details are right.
Then open Merchant Center and compare those answers with your feed. Every mismatch points to a data field to fix, and fixing it helps your Shopping ads at the same time.