Checkout used to be a page you built. A form, a “Buy” button, a confirmation screen you controlled pixel by pixel. That page is starting to disappear. Not because people stopped buying things online, but because an AI agent is increasingly the one filling out the form, and it never renders your checkout UI at all.

In September 2025, Stripe and OpenAI launched Instant Checkout inside ChatGPT. Four months later, Google and Shopify unveiled a competing standard, the Universal Commerce Protocol, at the National Retail Federation conference. Two open specifications, backed by different coalitions, both betting that checkout is becoming a protocol instead of a page. Here is what merchants need to understand about both, and what to actually do about it.

The shift in one sentence

In traditional ecommerce, the merchant designs the checkout experience. In agentic commerce, the agent does. ChatGPT or another assistant shows the shopper a product, a price and shipping options inside the chat itself; the shopper confirms, and the merchant’s system receives an API call instead of a page visit.

Stripe’s own agentic commerce documentation puts it bluntly: “The parts of commerce that used to be user experience problems are becoming protocol problems.” Instead of tuning button colors and form layouts, merchants are now defining API endpoints and product feeds. Emily Glassberg Sands, Stripe’s Head of Information and Data Science, framed the stakes more broadly: “Agents don’t just change who’s at the checkout. They change who’s doing the searching, the deciding, the trusting. All of it.”

This is not a future scenario. As of early 2026, ChatGPT Instant Checkout is live for U.S. users on Free, Plus and Pro plans, with Etsy, Instacart and Walmart already processing orders through it. Shopify’s Agentic Storefronts are on by default for eligible merchants, syndicating catalogs to ChatGPT, Google AI Mode, Microsoft Copilot and Perplexity at once. Perplexity added Instant Buy with PayPal in November 2025, connecting to Wayfair, Abercrombie & Fitch and thousands of BigCommerce and Wix stores. Even Anthropic, which has been more cautious about commerce, confirmed in February 2026 that it is building agentic purchasing into Claude, while committing to keep the experience free of sponsored placements.

Two protocols, two different bets

The Agentic Commerce Protocol (ACP) came first, announced September 29, 2025 by OpenAI and Stripe under an Apache 2.0 license. It defines a four-party model: the buyer approves, the agent presents products and handles the checkout interface, the merchant remains the merchant of record processing payment and fulfillment, and a payment service provider handles credentials securely. ACP can run as a REST API or as an MCP server, which is worth knowing if you’ve already looked at agent protocol infrastructure for other parts of your site.

Stripe made ACP practical to adopt with its Agentic Commerce Suite, launched December 11, 2025. Ahmed Gharib, Stripe’s Product Lead for Agentic Commerce, described it as “a low-code solution enabling businesses to sell across multiple AI agents via a single integration” — without it, Stripe estimates bespoke integration with each AI platform could take up to six months of engineering. The Suite covers catalog syndication, a checkout built on Stripe’s Checkout Sessions API, and payment processing via Shared Payment Tokens and Stripe Radar. Beyond OpenAI, Stripe lists Microsoft Copilot, Anthropic, Perplexity, Vercel and Replit as AI platform partners, with Squarespace, Wix, WooCommerce, BigCommerce and commercetools integrated on the merchant side.

The Universal Commerce Protocol (UCP) arrived January 11, 2026, co-developed by Shopify and Google and presented by Google CEO Sundar Pichai. Its co-developers include Etsy, Wayfair, Target and Walmart, with more than 20 companies endorsing it at launch, including Mastercard, Visa, Best Buy, Home Depot and, notably, Stripe itself. Where ACP is scoped tightly to checkout, UCP is built as a full commerce standard covering discovery through post-purchase, modeled architecturally after TCP/IP. It’s protocol-agnostic, supporting REST, MCP, A2A and Google’s Agent Payments Protocol (AP2), and merchants publish their capabilities at a discoverable /.well-known/ucp endpoint, similar to how A2A agents publish agent cards. Ashish Gupta, VP/GM of Merchant Shopping at Google, summed up the logic: “The shift to agentic commerce will require a shared language across the ecosystem.”

The two are not a fight merchants need to referee. Shopify’s Agentic Storefronts handle both simultaneously, meaning the same catalog can serve ChatGPT via ACP and Google AI Mode via UCP from one admin panel. Shopify VP of Product Vanessa Lee put it plainly: “Agentic commerce has so much potential to redefine shopping and we want to make sure it can scale.” For merchants weighing which specification actually matters for a smaller storefront, that question deserves its own closer look at agentic commerce for small merchants.

Solving trust when nobody’s at the checkout

Both protocols run into the same hard problem: traditional fraud checks assume the person holding the card details is the buyer. In agentic commerce, the agent has been authorized to act on the buyer’s behalf, but it is not the buyer. Stripe’s Kevin Miller wrote in an October 2025 post that “trust can’t be inferred. It has to be explicitly granted, scoped, and enforced in code.” Javelin Strategy & Research, cited by Visa, calls this the shift from “card-not-present” to “person-not-present” transactions.

Stripe’s answer is the Shared Payment Token (SPT): the buyer saves a payment method with an AI platform, and when a purchase is approved, the platform issues a token scoped to that merchant, capped at the checkout amount, and time-limited. The merchant creates a Stripe PaymentIntent from the token without ever seeing the buyer’s actual card number, and Stripe applies fraud detection in real time. Enabling SPTs on an existing Stripe integration takes, in Stripe’s words, “as little as one line of code.”

Card networks built their own layers on top. Visa’s Trusted Agent Protocol, launched October 2025 with Cloudflare, uses HTTP Message Signatures to help merchants tell legitimate agents apart from malicious bots. Mastercard’s Agent Pay, launched in April 2025, introduces “Agentic Tokens” built on existing tokenization rails, with U.S. issuer support live since November 2025 and global rollout underway. PayPal joined the ACP ecosystem on October 28, 2025, wiring PayPal wallets into ChatGPT checkout. Google’s parallel answer is AP2, announced September 2025 with more than 60 industry partners, which uses Verifiable Digital Credentials and a cryptographic “Mandate” system to prove user consent at each step, and, through a Coinbase x402 extension, even supports stablecoin payments.

Fraud detection itself had to change, because agents don’t leave the behavioral fingerprints humans do. Stripe built what it calls a foundation model for payments, a transformer trained on tens of billions of transactions that treats each charge as a token in a sequence. The risk isn’t theoretical: a June 2025 study found every ecommerce agent tested was vulnerable to visual prompt injection hidden in product listings, and research accepted to IEEE S&P 2026 found 13% of randomly sampled ecommerce sites had already exposed chatbot plugins to indirect injection through content like reviews. On the consumer side, Visa’s analysis of Javelin’s survey data found 88% of consumers are concerned AI will be used for identity fraud.

Where adoption actually stands

The projections vary enough to signal how early this still is. McKinsey projects $1 trillion in U.S. retail revenue orchestrated by agents by 2030, scaling to $3-5 trillion globally. Gartner expects AI agents to handle 90% of B2B purchases within three years, intermediating $15 trillion in spending by 2028. Forrester predicts a third of retail marketplace projects will be abandoned by 2026 as answer engines divert traffic away from them.

Consumers are more hesitant than the infrastructure suggests they should be. A Contentsquare survey found only 30% of 1,300 U.S. consumers willing to let an agent complete a purchase on their behalf. YouGov found 65% trust AI to compare prices but just 14% trust it to actually place an order, rising to 20% among Gen Z. Yet the traffic is already real: Adobe Analytics measured a 4,700% year-over-year jump in AI-driven traffic to U.S. retail sites by mid-2025, and Shopify reported AI-search-attributed orders grew 11x since January 2025. That gap between rising traffic and cautious buyers is worth tracking alongside the broader question of how brands compete in agentic commerce, since visibility and conversion are moving at different speeds.

Academic work is starting to describe what happens once agents do the choosing. A Columbia Business School and Yale study from August 2025 tested six frontier models, including Claude and GPT-4, and found AI shopping agents show “choice homogeneity,” concentrating demand on a narrow set of products with strong position bias in how listings are ranked. The researchers warn of winner-take-all dynamics and the rise of “AI-SEO,” where listings get optimized for agent behavior rather than human taste. That overlaps with the concerns raised around why most top retailers are still invisible to agentic commerce in the first place: agents can only concentrate demand on products they can actually parse.

What to do right now

For most merchants, the entry point is more accessible than the protocol names suggest. If you’re on Shopify, Agentic Storefronts may already be active; check the dashboard’s agentic channel settings and make sure your product data is complete. If you’re on Stripe, enabling Shared Payment Tokens is close to a one-line change, and the Agentic Commerce Suite handles the rest. BigCommerce, Wix, Squarespace and WooCommerce all have Stripe Suite integrations already available.

Beyond flipping the integration on, agents still need to be able to read your catalog. That means specific, descriptive product titles instead of vague ones, full descriptions covering materials and use cases, accurate real-time pricing and stock, and consistent categorization. Add Product schema with name, description, image, sku and brand, nested Offer schema with price and availability, and AggregateRating if you have reviews. This is exactly the machine-readable layer covered in our guide to preparing your site technically for Google’s UCP. Then test it: ask ChatGPT, Perplexity and Google AI Mode to recommend products in your category and see whether you show up. Finally, track agent referrals separately, since ChatGPT appends utm_source=chatgpt.com to outbound links, letting you watch conversion trends before they become significant revenue.

Walmart CEO Doug McMillon summed up the direction plainly: “For many years now, ecommerce shopping experiences have consisted of a search bar and a long list of item responses. That is about to change.” Whether that change reaches your revenue this year or next depends less on which protocol wins and more on whether your product data is ready when an agent comes looking for it.

Frequently asked questions

What is agentic commerce?

Agentic commerce is the model where an AI agent — inside ChatGPT, Google AI Mode, Perplexity or similar — discovers products, compares options and completes a purchase on a shopper’s behalf, using structured APIs instead of a rendered checkout page.

Do I need to choose between ACP and UCP?

No. The two protocols solve overlapping problems from different angles, and platforms like Shopify’s Agentic Storefronts already support both at once, syndicating the same catalog to ChatGPT via ACP and Google AI Mode via UCP.

How do agent payments stay secure without exposing card details?

Through scoped, time-limited tokens such as Stripe’s Shared Payment Tokens, paired with network-level frameworks like Visa’s Trusted Agent Protocol and Mastercard’s Agent Pay, all designed to verify that a purchase is a legitimate agent action rather than fraud.

Is agentic commerce actually generating sales yet?

Yes, though modestly relative to total ecommerce. AI-referred traffic to U.S. retail sites grew 4,700% year-over-year by mid-2025, and Shopify reported an 11x increase in AI-search-attributed orders since January 2025, even though only 14% of consumers currently trust AI to place an order for them.