Google’s AP2, Stablecoins—and the Secret to Being “Agent-Readable”

The Sale You Never Saw

Picture this: a shopper is looking for a winter jacket. She doesn’t just want “something warm.” She wants a specific color—forest green—delivered within five days, under $180. So, she tells her AI assistant exactly that.

In seconds, her assistant scans online stores across countries, checks product options, shipping timelines, and return policies. Then it does something new: it creates a digitally signed instruction—a “mandate”—that says, in effect, “If you find a forest green jacket under $180 that arrives within five days, buy it.”

The agent finds two matching offers.

The first store is beautifully designed, with glossy lifestyle photos and a slick homepage. But under the surface, it's a bit of a black box. The AI can’t clearly verify the shipping date, check if the green color is truly in stock, or understand the return policy. The discount is only applied at checkout, so the final price isn’t clear upfront.

The second store looks plain, maybe even a little boring. But everything is labeled in a way machines understand: shipping timelines by region, exact product variants, a clearly stated return policy, and real-time inventory. The AI can verify every detail, line by line. It signs the mandate and completes the purchase—without a human ever seeing a checkout page.

The buyer didn’t “choose” Merchant B. Her assistant did.

And the reason it chose Merchant B had nothing to do with branding. It had everything to do with clarity.

From Browsing to Buying—in a Blink

Welcome to the emerging world of agentic commerce—where AI assistants go beyond product recommendations, and instead complete purchases based on specific conditions you give them.

Google’s newly announced Agent Payments Protocol (AP2) is one of the clearest signals yet that we’re heading into a new phase of online commerce—one where agents can act on behalf of their users with precision and trust.

It builds on what Mastercard, Coinbase, and others have already been exploring: how to make payments and intent programmable. In this new world, your customer’s journey doesn’t stretch across ten clicks and five funnels. The discovery, decision, and purchase can collapse into a single, fluid moment—if the store can speak the agent’s language.

And that’s the key: being readable. Not just for people. For agents.

The New Visibility: Being Clear to Machines

We’re all familiar with SEO—the art of making your site readable to search engines. The next big shift is similar, but broader: making your store readable to agents that don’t just find, but buy.

It starts with exposing the basics in a way machines can trust. Think product details, prices, shipping policies, and inventory—not buried in dropdowns or dynamic scripts, but cleanly marked up so they can be verified before a purchase is committed.

For instance, instead of showing “Fast Shipping!” in flashy type, agents need to see: “Ships to Germany in 3 business days via DHL Express, for €9.90.” Unlike humans, these agents aren’t vulnerable to slogans but rather hard signals.

Similarly, instead of styling color choices as clickable images, call the color “forest” in your structured data. Use standard formats (like Schema.org markup) so the agent can reliably read and act on the information.

The more precise your store’s data, the more likely an agent will trust it—and commit to a transaction.

Stablecoins and Rail Choice: Why It Matters

Agents don’t just look at your products. They look at how you're set up to get paid.

Google’s AP2 protocol is rail-agnostic, which means agents can choose between traditional cards, real-time payments (like SEPA Instant or FedNow), and increasingly, stablecoins—cryptocurrencies that hold a stable value and enable fast, borderless transactions.

If your store only supports one type of payment method, you may fall off the shortlist. But if your PSP (payment service provider) supports multiple rails, including stablecoins via trusted platforms like Coinbase Commerce, agents will factor that flexibility into their decisions.

This is especially important for global sellers. An agent buying on behalf of a customer in São Paulo may prefer Pix or USDC to avoid weekend banking delays and currency conversion costs. The agent doesn’t need to argue with your checkout. It simply moves on to a merchant that offers cleaner, faster rails.

How to Get Started—Without Getting Overwhelmed

The good news? You don’t need a team of engineers or a deep understanding of AI. What you need is consistency and clarity.

Start by:

• Naming things clearly. Use real product attributes in your markup—colors like “forest,” sizes like “M,” and delivery windows like “5 business days.”

• Structuring your key policies. Spell out returns, shipping, and price conditions in formats that machines can read—not just in long-form prose or hidden behind modals.

• Exposing what you know. If your jacket is out of stock in large, say so. If a promo price applies, show the net price directly. Don’t hide crucial info behind scripts or dynamic checkout flows.

• Providing choices. Offer more than just cards. The more payment options you support—especially those that are programmable—the more likely an agent will complete the purchase.

• Standardizing URLs and anchors. A link like /returns#30-day-window is easier for AI to quote and link than a generic or broken URL.

You don’t have to do it all at once. But the sooner your site becomes “agent-readable,” the more discoverable—and trusted—you’ll be in this new landscape.

B2B: Where Agents Take the Friction Out

For businesses that sell wholesale, or operate across long supply chains, the benefits are even more dramatic.

Imagine your buyer’s AI assistant requesting net-30 terms, submitting a PO, and completing a five-figure order in minutes—all while complying with your terms. No email chains, no back-and-forth. Just structured data doing what it’s supposed to do.

Solutions like TreviPay will play a big role here. If your site or PSP exposes net terms options, agents can factor that into their purchase conditions—just like they would with shipping or pricing.

This means you can capture larger, recurring B2B orders with less overhead, fewer manual steps, and faster cash flow.

Fraud, Liability, and the Future of Trust

Of course, the question arises: what if the agent gets it wrong?

The evolving agent payment protocols like AP2 include secure signatures, logs, and proof of intent. That means you’ll know who authorized what—and when. Merchants will be able to track agent-initiated orders, flag them as a distinct category, and apply specific risk rules.

Fraud detection isn’t going away but it is getting smarter.

The Quiet Race Has Begun

Google’s AP2 launch—developed alongside Mastercard, PayPal, Coinbase, and Shopify—marks a shift in how transactions are initiated, verified, and completed. 

What used to happen across a landing page, a checkout funnel, and a confirmation screen now collapses into a machine-to-machine interaction. Your buyer might never visit your site directly, but their assistant will parse what you’ve published, check it against a mandate, and decide in milliseconds.

When the inputs are clean, the decisions are instant. That’s where the advantage now is. 

So if you're thinking about how to future-proof your store for this new world, start with one question:

Can an agent understand you?