ObsurfableAI shopping used to mean "ChatGPT recommends a product and the user Googles it." In 2026, the recommendation and the purchase happen in the same conversation.
Two protocols now define how that works: ACP (Agentic Commerce Protocol) from OpenAI and Stripe, and UCP (Universal Commerce Protocol) from Google and a coalition of retailers. They are not competitors in the sense of winner-take-all — they are parallel rails into different AI surfaces, and most brands will need both.
This extends the product-discovery layer we covered in How Amazon Rufus and ChatGPT Shopping Decide What to Recommend — but at the infrastructure level.
The protocol landscape
| Protocol | Owners | Launched | Primary surface | Focus |
|---|---|---|---|---|
| ACP | OpenAI + Stripe | Sept 2025 | ChatGPT | Instant checkout in chat |
| UCP | Google + Shopify coalition | Jan 2026 | Google AI Mode, Gemini | Full shopping lifecycle |
| MCP | Anthropic | 2024 | All (transport layer) | Agent-to-data connections |
| AP2 | Google coalition | 2026 | Cross-platform | Secure agent payments |
MCP (Model Context Protocol) is the underlying transport — it lets AI agents query product catalogs, check inventory, and access tools in real time. ACP and UCP are commerce-specific protocols built on top of (or alongside) MCP.
ACP: buy inside ChatGPT
ACP powers ChatGPT's "Buy" button and Instant Checkout. The flow:
- User asks ChatGPT a shopping question
- ChatGPT searches merchant product feeds
- Products appear with recommendations and a "Buy" button
- User clicks Buy — checkout happens in-chat via Stripe's Shared Payment Token
- Merchant remains merchant of record; the agent never sees payment credentials
Merchants need: a product feed, a REST Checkout API (create, update, get, complete, cancel session endpoints), and Stripe integration.
ACP is live. It is ChatGPT-first and optimized for conversational product discovery.
UCP: Google's open commerce standard
UCP, co-developed by Google with Shopify, Etsy, Wayfair, Target, and 20+ partners, standardizes the entire shopping lifecycle for AI agents:
- Product discovery and comparison
- Cart management
- Checkout and payment (via AP2)
- Order tracking and post-purchase support
Merchants publish a capability profile at /.well-known/ucp so any UCP-compliant agent can discover what they support.
UCP is platform-agnostic by design — it works across Google AI Mode, Gemini, and any future UCP-compliant agent. Shopify merchants can enable it with a toggle in admin.
Why visibility precedes the transaction
Both protocols solve the checkout problem. Neither solves the discovery problem — and discovery is where AEO lives.
Before a user clicks "Buy" in ChatGPT or asks Google's AI to compare products, the agent has to know your product exists and trust your brand enough to recommend it. That decision is made the same way all AI recommendations are made:
- Structured product data models can parse
- Third-party validation (reviews, roundups, Reddit discussions)
- Brand entity clarity across the web
- Fresh, extractable comparison content
A merchant with perfect UCP compliance but zero AI visibility will process transactions for products nobody asks about.
What brands should do now
1. Implement both protocols if you sell online.
Shopify stores: enable AI channels in admin (UCP is native; ACP via Stripe integration). Custom storefronts: implement the required endpoints — UCP capability profile and ACP checkout API.
2. Optimize product feeds for AI extraction.
Clear product names, structured attributes, pricing, availability, and comparison-friendly descriptions. Agents query feeds in real time — messy data means invisible products.
3. Monitor AI recommendations separately from protocol compliance.
Protocol implementation gets you into the transaction layer. AEO gets you into the recommendation layer. You need both.
4. Earn third-party validation.
Listicles account for 63% of LLM citations. Product reviews on G2, Amazon, and editorial roundups shape what agents recommend before checkout is even possible.
5. Track agentic commerce as a channel.
Only ~16% of brands track AI visibility at all (industry estimates). First movers on both protocol compliance and recommendation visibility capture disproportionate agentic traffic.
How Obsurfable fits
Obsurfable monitors whether your products and brand appear when users ask AI assistants shopping questions — the layer upstream of any checkout protocol. Define prompts like "best [category] tool" or "where to buy [product type]" and track whether you are recommended, cited, or absent.
The bottom line
UCP and ACP are the plumbing for agentic commerce. AEO is the marketing. Brands that implement protocols without optimizing for AI recommendations will have checkout ready for a customer who never hears their name.