Guide
Agentic commerce
Agentic commerce is online shopping carried out by AI agents on a buyer's behalf — the agent discovers products, compares them, and can complete checkout inside an assistant like ChatGPT, Google's AI Mode, or Perplexity, instead of on your storefront. For a merchant it raises one blunt question: can an AI agent actually complete a purchase on your store? Most tools only check whether you get mentioned; readiness is about whether you can be bought from. A store is ready when its catalog, structured data, discovery files, and checkout endpoints satisfy the protocols agents read — chiefly the Agentic Commerce Protocol (ACP) and the Universal Commerce Protocol (UCP). This guide explains those protocols, the five AI shopping surfaces, and exactly what to fix.
Agentic commerce vs. conversational commerce vs. SEO
Conversational commerce is a human chatting with a store (live chat, messaging). Traditional SEO and AI search optimize for being found and mentioned. Agentic commerce is different: an AI agent acts for the buyer — discovering, comparing, and transacting — often with no human in the storefront at all. The work that makes you legible to agents overlaps with structured-data SEO, but the goal is transactability, not a ranking.
The protocols agents read
Four standards have settled into a layered stack. They are complementary, not competing — each handles a distinct concern.
ACP — OpenAI and Stripe's protocol behind ChatGPT shopping. ChatGPT uses a merchant-push feed (products are submitted, not crawled), plus an agentic-checkout API and a delegated-payment flow. Shopify stores are auto-enrolled; other platforms apply and build a feed. In March 2026 OpenAI pivoted away from in-chat Instant Checkout toward handing off to the merchant's storefront — a sign that checkout at scale is harder than the launch implied.
UCP — Google and Shopify's protocol for the full lifecycle (discovery → checkout → payment → post-purchase), behind Google AI Mode and Gemini. Merchants declare capabilities at /.well-known/ucp. It's transport-agnostic and charges no merchant fee in AI Mode.
MCP and A2A — the plumbing beneath them: MCP connects agents to tools (search a catalog, create a cart); A2A lets a buyer's agent talk to a merchant's agent. AP2 carries cryptographic proof of payment authorization and plugs into UCP.
The five AI shopping surfaces
We score readiness across the five surfaces an agent can shop on. Each reads slightly different signals, so being ready for one does not make you ready for the others.
- ChatGPT (ACP)
- Runs on a merchant-push ACP feed: products are submitted, not crawled. Shopify and Etsy stores are auto-enrolled; other platforms apply at
chatgpt.com/merchantsand build a feed. OpenAI charges a fee on completed purchases. - Google AI Mode (UCP)
- Reads your UCP profile at
/.well-known/ucpalongside your Merchant Center feed, with no added merchant fee. Same protocol powers Gemini shopping. - Perplexity
- The one surface that crawls your pages directly and reads schema.org/Product JSON-LD and GTINs off the page, so on-page structured data has the most direct impact here. Its merchant program is free to join.
- Microsoft Copilot
- Aligns to UCP and leans on Bing and Merchant Center signals. Merchant-facing documentation is thin, so parts of this surface are inferred; we mark those as inferred in the report rather than assert them.
- Meta AI
- The newest of the five, turning social discovery on Instagram and Facebook into agent-assisted shopping. Capabilities are still landing through 2026.
What a merchant must do — fix the data layer
The highest-ROI work is data hygiene, because it pays off on every surface at once and regardless of which protocol wins: complete schema.org/Product JSON-LD on every product page, GTINs on every product that has one, accurate real-time pricing and availability, crawlable shipping and return policies, AI crawlers allowed in robots.txt, and server-rendered pages (AI bots don't run JavaScript). An /llms.txt manifest is a cheap positive signal on top.
We turn that into a concrete, ranked checklist: see all 82 checks, or read the platform specifics for BigCommerce, WooCommerce, and Shopify. Weighing two platforms? See BigCommerce vs Shopify for AI commerce.
Where it stands in 2026
The opportunity is real but early. AI-referred shopping sessions grew more than 8× year-over-year on Shopify and orders from them about 13× (Shopify, Q1 2026, via PYMNTS), and generative-AI traffic to retail sites is up several thousand percent off a tiny base (Adobe Digital Insights). Yet AI still drives on the order of ~1% of retail web traffic today, and even on Shopify's default-on infrastructure, OpenAI's checkout pivot showed agent transactions are still being figured out. The trajectory still points up: Morgan Stanley estimates $190–385B of US e-commerce spend influenced by agentic shoppers by 2030, and Gartner projects 90% of B2B buying will be agent-intermediated by 2028. That's why getting the data layer right now, while it's cheap, compounds.
Can an AI agent buy from your store?
Score your store 0–100 across all five surfaces and get a ranked, deep-linked fix list — free. For the audit-specific walkthrough, see agentic commerce readiness.
FAQ
- What is agentic commerce?
- Agentic commerce is online shopping carried out by an AI agent on a buyer's behalf — the agent discovers products, compares them, and can complete checkout inside an assistant like ChatGPT, Google AI Mode, or Perplexity, instead of the buyer browsing your storefront.
- What's the difference between ACP and UCP?
- ACP (OpenAI/Stripe) powers ChatGPT shopping via a merchant-push feed and a checkout API. UCP (Google/Shopify) covers the full lifecycle for Google AI Mode and Gemini and is declared at /.well-known/ucp. Being ready for one does not make you ready for the other — they read different signals.
- Is agentic commerce just SEO or AI search?
- No. Traditional SEO and AI search optimize for being found and mentioned. Agentic commerce is about whether an agent can understand your products and complete a purchase — structured product data, crawlable policies, and the checkout signals the protocols require. Most tools check whether you get mentioned; readiness is whether you can be bought from.
- What should a small merchant do first?
- Fix the data layer: complete schema.org/Product JSON-LD on every product page, add GTINs, keep pricing and availability accurate, and don't block AI crawlers in robots.txt. It's low-cost, pays off on every surface at once, and matters regardless of which protocol wins.
- Does my platform handle this automatically?
- Partly, and it varies. Shopify auto-enrolls stores into ChatGPT and ships UCP/MCP natively. BigCommerce supports UCP and agent checkout through marketplace integrations (e.g. PayPal Store Sync), not a default-on toggle. WooCommerce reaches agent checkout via Stripe's Agentic Commerce Suite. On any platform, your product data still has to be good enough.
- How do I check whether an agent can buy from my store?
- Run a free audit. We score a store 0–100 across 82 checks and all five AI shopping surfaces against the live ACP and UCP specs, and return a ranked, deep-linked fix list. It reads only public signals and doesn't modify the store.