For twenty years, ecommerce marketing has been about persuading a person scrolling a results page. Agentic commerce adds a new decision-maker: an AI agent that reads product data, compares options and shortlists, or even buys, on the shopper’s behalf. If the agent cannot understand your products, it will not choose them.
This is no longer theoretical. In 2026 Google launched the Universal Commerce Protocol and agentic checkout in AI Mode, OpenAI published the Agentic Commerce Protocol, and both companies have adjusted their plans as adoption has played out. The details keep changing, but the fundamentals of being chosen are becoming clear.
This guide explains how AI shopping agents choose products, what is actually live, and a step-by-step plan to make your catalogue agent-ready. For background on the wider trend, read our explainer on agentic search.
Key Takeaways
- AI agents choose products from structured, accurate, consistent data, so your product feed is now a core marketing asset.
- Google’s UCP and OpenAI’s ACP let agents complete purchases while the retailer stays merchant of record.
- Rollouts are US-first and still evolving; OpenAI moved Instant Checkout into apps in March 2026.
- Reviews, clear differentiation and consistent facts across the web help agents trust and recommend you.
- You can prepare now without joining any protocol: fix feeds, schema, policies and site accessibility.
How AI Shopping Agents Choose Products
Agents do not browse the way people do. They pull structured data, read pages, check prices and availability, weigh reviews and match all of that against the shopper’s request and preferences. OpenAI’s Help Center describes ChatGPT surfacing items it judges relevant to the user’s intent, using the query and context such as Memory, and factors like price and reviews can also weigh in. It is explicit that these are not paid placements:
“Product results are selected independently by ChatGPT and are not ads, nor influenced by any OpenAI partnerships.”
— OpenAI Help Center, Shopping with ChatGPT search
Google is moving in the same direction with richer product data. In January 2026 it announced dozens of new Merchant Center attributes designed for conversational shopping, such as answers to common product questions, compatible accessories and substitutes. Our post on how AI chooses sources covers the content side of this selection process.
What Is Live in 2026
| Initiative | What it does | Status (October 2026) |
|---|---|---|
| Google Universal Commerce Protocol (UCP) | Open standard for agents to discover, buy and handle post-purchase; powers checkout on eligible listings in AI Mode and the Gemini app | Launched January 2026; US first, with Canada, Australia and later the UK announced at GML 2026 |
| Google Business Agent | Branded AI agent that answers shopper questions on Search in the retailer’s voice | Live for eligible US retailers via Merchant Center |
| Google Direct Offers | Lets advertisers show exclusive offers to shoppers ready to buy in AI Mode | Pilot since January 2026, expanding |
| OpenAI Agentic Commerce Protocol (ACP) | Product feeds, checkout sessions and delegated payment between ChatGPT and merchants | Open to build with; Instant Checkout for approved partners |
| ChatGPT shopping | Product results in ChatGPT search | Live; in March 2026 OpenAI moved checkout into merchant apps within ChatGPT |
The OpenAI shift is a useful reminder not to over-invest in any single checkout surface. In March 2026, according to Shopifreaks’ summary of reporting by The Information, OpenAI said Instant Checkout was moving to apps, with shoppers paying through a retailer’s app or being sent to the retailer’s site.
Who Owns the Customer?
A common worry is losing the customer relationship to the AI platform. Both main protocols keep the retailer in charge of the sale. Google’s UCP FAQ states plainly:
“You are the Merchant of Record.”
— Google for Developers, Universal Commerce Protocol FAQ
OpenAI’s developer documentation similarly says OpenAI is not the merchant of record in the Agentic Commerce Protocol; merchants use their own payment provider. You still need to fulfil orders, handle returns and provide support, so operational quality directly affects whether agents keep recommending you.
How to Get Your Products Chosen: Step by Step
- Audit your product feed. Make titles descriptive, fill every relevant attribute (size, material, compatibility, GTIN), and keep price and availability in sync with your site. Our product feed optimization guide covers the detail.
- Complete Merchant Center. Google’s UCP FAQ says product listings, branding, return terms and business contact details must be complete and current. Start with our Google Merchant Center guide.
- Add structured data to product pages. Product, offer, shipping and return policy markup help machines read your pages; see our guide to product schema and merchant listings.
- Answer the questions agents will ask. Add clear specs, use cases, comparisons and FAQs on product pages, written in plain language.
- Build review volume and quality. Agents weigh reputation. Our article on product reviews and SEO explains how to collect and show them.
- Keep facts consistent everywhere. Prices, policies and specs should match across your site, marketplaces and feeds.
- Make your site agent-friendly. Ensure key pages are crawlable, fast and usable without complex scripts; see our guide to optimizing for AI agents and the emerging WebMCP standard.
- Evaluate protocol participation. If you are on Shopify or another supported platform, check which agentic integrations are available; larger retailers can apply directly.
Agent Readiness Checklist
| Area | What agents need | Quick check |
|---|---|---|
| Product identity | Unique IDs, GTINs, brand, model | Feed diagnostics show no missing identifiers |
| Attributes | Size, colour, material, compatibility | Top 50 products have every relevant attribute filled |
| Price and stock | Real-time, consistent values | No price or availability mismatches in Merchant Center |
| Policies | Shipping costs, delivery times, returns | Configured in Merchant Center and marked up on site |
| Trust | Reviews, ratings, brand reputation | Recent reviews on key products |
| Content | Clear descriptions and answers | Product pages answer the top five customer questions |
Common Mistakes
- Treating the feed as an afterthought. Thin or inaccurate feeds are the fastest way to be ignored by agents.
- Chasing a single platform. Checkout surfaces have already changed; build data foundations that work everywhere.
- Inconsistent pricing. Agents compare sources, and mismatches undermine trust.
- Marketing copy without facts. "Best quality ever" helps nobody; dimensions, materials and compatibility do.
- Ignoring post-purchase. Returns and support affect reviews, and reviews affect future recommendations.
- Blocking useful crawlers by accident. Check robots.txt and bot protection so shopping-related crawlers can reach product pages you want recommended.
Related Guides
- Product Bundling: Strategies That Raise Order Value
- Free Shipping Strategy: Thresholds, Costs and Conversions
- Product Photography for Ecommerce: A DIY Guide
Frequently Asked Questions
What is agentic commerce?
Agentic commerce is shopping where an AI agent does much of the work for the customer: understanding the need, finding and comparing products, and in some cases completing checkout through a protocol that connects the agent to the merchant.
Do I need to join a protocol like UCP or ACP to be recommended?
No. Products can be recommended in AI answers if your product data is accessible, for example through your website, Merchant Center or a platform integration. Protocols matter for completing the purchase inside the AI experience rather than sending the shopper to your site.
Who is the merchant of record in agentic checkout?
In both Google's Universal Commerce Protocol and OpenAI's Agentic Commerce Protocol, the retailer remains the merchant of record and processes payment through its own payment provider.
Is agentic checkout available outside the US?
Launches have been US-first. At Google Marketing Live 2026, Google said many UCP features would roll out in the US first, then Canada and Australia, with the UK later. Check each platform's current documentation for your market.
Are product recommendations in ChatGPT paid ads?
According to OpenAI's Help Center, product results in ChatGPT search are selected independently and are not ads. Visibility depends on relevance, product data and signals such as price and reviews.
Conclusion
Agentic commerce rewards the brands that are easiest for machines to understand and trust: complete product data, accurate prices and stock, clear policies and genuine reviews. The protocols will keep evolving, but those foundations will not. If you want help preparing your catalogue for AI shopping, our ecommerce SEO services cover feeds, schema and content, and you can request a custom quote today.
References
- Google: New tech and tools for retailers to succeed in an agentic shopping era
- Universal Commerce Protocol FAQ – Google for Developers
- Key concepts – Agentic Commerce, OpenAI Developers
- Shopping with ChatGPT search – OpenAI Help Center
- Shopifreaks: OpenAI scales back its integrated commerce plans
- Google Marketing Live 2026: Ask Advisor, AI Mode ads and UCP – TechWyse
- Semrush: Agentic Commerce Is Here: What It Means for the Ecommerce Industry
