AI Shopping Is Getting Real: Here's What Online Stores Need to Be Ready

What changed

Mastercard announced Agent Connect on September 9, meant to give merchants a single integration through which participating AI shopping agents can discover products and current availability, recommend items, confirm final price, tax, shipping and fulfillment, and complete an authorized purchase. Merchants keep control over which AI experiences they participate in, along with pricing, fulfillment, customer relationships and their own business rules. The rollout starts in the U.S. Mastercard says it will also help merchants prepare their product information specifically for AI discovery and make catalogs available to third-party agents, and it's incorporating Anthropic's recently announced commerce-agent blueprint. Yesterday's piece covered Meta's Muse agent being able to browse and pay for a customer. This is the next piece: a payment network building the plumbing that connects a merchant's own catalog to whichever agents show up wanting to shop it.

The honest nuance

The infrastructure is arriving faster than consumer trust in it. New Visa research released September 9 found that while 72% of surveyed consumers have used an AI assistant, only 23% trust generative AI to handle payments for them on its own. That trust rises to 61% when a payment network like Visa is involved in the transaction. That gap is worth sitting with: shoppers are far more comfortable letting an AI agent buy something when a company they already trust with their card is standing behind it. It suggests payment networks may end up playing the same trust role for AI agents that they've long played for online commerce generally, verifying and backstopping a transaction between two parties who don't fully trust each other yet.

Why it matters for your business

The pieces being covered separately over the past two weeks are starting to connect into something closer to an actual system: AI product discovery, merchant catalogs, agent recommendations, authorized checkout, and payment networks, rather than a handful of unrelated demos. You don't need to integrate Mastercard Agent Connect tomorrow, and most small stores won't be first in line for it. The more useful signal is what Mastercard is choosing to build infrastructure around, because it's a concrete answer to a question that's been vague until now: what does "AI-ready" actually mean for an online store? It means an AI shopper needs to reliably retrieve what the product is, whether it's in stock, what it costs, what shipping adds, what the fulfillment terms are, and whether checkout can actually complete without a human on either end.

What to actually do

The practical work here is the same boring, valuable work as always, just with a clearer reason to prioritize it: accurate and complete product catalogs, descriptive product pages, structured product data (schema.org markup, not just pretty formatting), current inventory counts, explicit shipping and return terms, and a checkout flow that doesn't quietly assume a human is the one filling it out. Test that last part yourself. Try to complete a purchase on your own site using only the information on the page, no calling support, no guessing at a policy that isn't written down. Anywhere you get stuck is exactly where an AI agent, and likely a fair number of human customers too, would also get stuck.

From Kindloom Labs

That same idea, making your page legible to software instead of just people, is also what's driving Google's AI Overviews. If you want the on-page basics for that side of it too, we covered it in What AI Search Means for Small Business Websites which walks through writing direct, quotable answers into your own pages.

Sources

• Mastercard gives merchants a simpler way to build, connect and scale with AI agents (Mastercard)
• New Visa Research Finds Consumer Trust Is Accelerating the Path to Agentic Commerce (Visa Investor Relations)

Keep reading