Agentic commerce is shopping where an AI agent, rather than a person clicking through pages, finds products, compares them and in some cases completes the purchase for the user. The Universal Commerce Protocol, UCP, is an open standard announced by Google with retail and payments partners to let agents talk to merchants in a consistent way. For marketers, the shift means product data, policies and checkout must be readable by machines, not only attractive to people.
What is agentic commerce?
In a classic online purchase, the shopper searches, opens product pages, reads reviews and checks out. In agentic commerce, the shopper gives an instruction such as “find running shoes under a set budget in my size, with free returns, and buy the best-rated pair”, and an AI agent carries out the steps. The person approves key moments, typically the final payment.
Several companies are building this. OpenAI introduced Instant Checkout in ChatGPT with the Agentic Commerce Protocol developed with Stripe. Google has added agentic checkout features to AI Mode in Search and the Gemini app. Perplexity, Amazon and payment networks such as Visa and Mastercard have announced agent-related shopping and payment programmes. The details move quickly, so treat any feature list as a snapshot.
What is the Universal Commerce Protocol?
UCP is Google’s proposed open standard for agent-to-merchant commerce, developed with partners including retailers, commerce platforms and payment providers. It describes how an agent discovers a merchant’s capabilities, reads product and offer data, builds a cart, applies policies and completes checkout with secure payment tokens, without custom integration for every store.
The idea is similar to how payment and feed standards made earlier commerce scale. If every agent needed a bespoke connection to every shop, agentic commerce would stay limited to a few big retailers. A shared protocol lets platforms such as Shopify or commerce software providers implement it once for many merchants. Adoption and exact specifications are still developing at the time of writing, so check the official documentation before planning engineering work.
Why should marketers care now?
Because the work that prepares a store for agents overlaps heavily with work that already pays off. Agents rely on structured, accurate data: product attributes, prices, stock, shipping times and return policies. Stores that keep that data clean also perform better in Google Shopping, marketplaces and site search today.
There is also a visibility question. If an agent chooses three products to show a user, the brands that are not chosen may not be seen at all. Being chosen depends on data completeness, price, reviews and policies an agent can read and compare. Clever ad copy matters less to a machine than a clear return window. Brand still matters, though: agents often respect a user’s stated preferences, so brands that people ask for by name keep an advantage.
How do AI agents choose products?
No vendor publishes a full ranking method, and claims of guaranteed placement should be doubted. Agents generally work from information they can retrieve and verify, which in practice includes:
- Product feeds and structured data with complete attributes such as size, material and compatibility.
- Real-time price and stock information.
- Shipping speed, cost and delivery areas.
- Return and warranty policies stated clearly.
- Review volume and themes from trusted sources.
- Merchant reliability signals, such as order defect rates on marketplaces.
These are the same facts a careful human shopper checks, just read in bulk and compared quickly.
Where does agentic commerce fit in 2026?
For most brands it is an emerging channel, not a main one. Volumes through agent checkouts are small compared with normal web and app shopping, and many users still prefer to browse. It fits best today in replenishment purchases, well-defined products with clear specs, and categories where comparison is tedious, such as electronics accessories or household supplies.
It fits less well for products bought on emotion or style, high-consideration purchases with long research cycles and anything needing fitting or consultation. Marketers should prepare foundations now and expand as adoption becomes measurable.
Measurement is still immature
Attribution for agent-driven orders is limited. Some platforms pass referral information; others do not. Tag orders by source where possible, watch referral traffic from AI platforms in analytics, and keep a log of which agent programmes your store has joined, so changes in sales can be traced.
What should a store do to prepare?
Preparation is mostly data and policy work:
- Audit the product feed for missing attributes, GTINs and accurate stock.
- Publish shipping and return policies in clear, structured form on the site and in Merchant Center.
- Add Product, Offer and MerchantReturnPolicy structured data to product pages.
- Check if your commerce platform supports agent checkout programmes or UCP, and what opting in requires.
- Review fraud and payment settings with your payment provider for agent-initiated orders.
- Make sure customer service can identify and handle orders placed through agents.
How Alien Road applies agentic commerce and UCP
In a typical engagement we start with a machine-readability audit. We check the client’s product feed, structured data, policy pages and Merchant Center settings for what an agent would see, and list gaps such as missing attributes or return policies buried in long legal text.
Then we fix the feed and markup, rewrite key policies into short structured statements and review platform options with the client’s development team, including whether their commerce platform offers agent checkout programmes. We set up analytics segments for AI referral traffic and order tagging where the platform passes source data. Reports track feed health, AI referral sessions and any agent-attributed orders, so the client can see the channel’s real size before investing further.
Where to start this week
Pick five products and read their data as an agent would: open the Merchant Center listing, the structured data test and the return policy page. Note every fact a shopper would ask about that is missing or unclear. Fixing those gaps helps Google Shopping today and agents tomorrow, so the effort is not wasted whatever pace adoption takes. Repeat the check for your top sellers next.