Selling To AI: The Complete Guide To Agentic Commerce

For 30 years, checkout has been a web page. A type with fields for identify, handle, bank card quantity. Whether or not it was Amazon’s one-click patent or Apple Pay’s fingerprint, the innovation was at all times about making that type quicker to get by way of.

The shape itself by no means went away. Now it’s.

That is the ultimate article in a five-part sequence on optimizing web sites for the agentic net. Part 1 lined the evolution from Web optimization to AAIO. Part 2 defined the right way to get your content material cited in AI responses. Part 3 mapped the protocols forming the infrastructure layer. Part 4 bought technical with how AI brokers understand your web site. This text covers the commerce layer: how AI brokers discover merchandise, full purchases, and deal with funds with out ever loading a checkout web page.

In September 2025, Stripe and OpenAI launched Instant Checkout inside ChatGPT. In January 2026, Google and Shopify unveiled the Universal Commerce Protocol on the Nationwide Retail Federation convention. Two open requirements. Two competing visions for a similar shift: checkout changing into a protocol, not a web page.

All through this text, we draw solely from official documentation, analysis papers, and bulletins from the businesses constructing this infrastructure.

How We Bought Right here

Each technology of commerce know-how has solved the identical drawback: decreasing the friction between “I would like one thing” and “I’ve it.” Agentic commerce isn’t a break from this sample. It’s the sample’s logical conclusion.

1994: The primary on-line buy. On Aug. 11, 1994, Phil Brandenberger used his bank card to purchase Sting’s Ten Summoner’s Tales CD for $12.48 from an internet site known as NetMarket. The New York Occasions lined it the subsequent day. NetMarket’s 21-year-old CEO, Daniel Kohn, informed the paper: “Even when the N.S.A. was listening in, they couldn’t get his bank card quantity.” Netscape’s SSL protocol, launched that very same 12 months, made it attainable.

The friction eliminated: You not needed to go to a bodily retailer.

Late Nineties: Comparability buying. Inside a couple of years, web sites like BizRate (1996), mySimon (1998), and PriceGrabber (1999) let patrons see costs throughout a number of retailers immediately. Google entered the house in 2002 with Froogle, later renamed Google Product Search in 2007, then Google Buying in 2012.

The friction eliminated: You not needed to go to every retailer to match.

1998: The shop adapts to you. Amazon deployed item-to-item collaborative filtering at scale, the algorithm behind “prospects who purchased this additionally purchased.” Greg Linden, Brent Smith, and Jeremy York revealed the underlying analysis in IEEE Internet Computing in 2003. In 2017, the journal named it the perfect paper in its 20-year historical past.

The friction eliminated: You not needed to know precisely what you wished.

2015: Commerce strikes into conversations. Chris Messina, then Developer Expertise Lead at Uber, coined the time period “conversational commerce” in a January 2015 Medium publish, describing “delivering comfort, personalization, and choice help whereas persons are on the go.” In April 2016, Mark Zuckerberg launched the Facebook Messenger Platform, declaring: “I’ve by no means met anybody who likes calling a enterprise.” In the meantime, in China, WeChat had already proved the mannequin. Its Mini Packages, launched January 2017, generated 800 billion yuan (~$115 billion) in transactions by 2019.

The friction eliminated: You not needed to open a retailer’s web site.

2014-2023: Voice and social commerce. Amazon Echo launched in November 2014, promising you possibly can purchase issues with no display. The promise was principally unfulfilled. Social commerce had higher luck: TikTok Store, launched within the U.S. in September 2023, reached $33.2 billion in global sales by 2024. Content material turned the storefront.

The friction eliminated: Buy intent was created contained in the feed, not looked for.

2024: AI begins purchasing for you. Inside months, each main platform launched AI buying options. Amazon launched Rufus in February, a conversational assistant skilled on its product catalog. Google rebuilt Shopping with AI in October, drawing on 50 billion product listings. Perplexity launched “Buy with Pro” in November, turning a search engine right into a retailer.

The friction eliminated: AI did the analysis, comparability, and advice for you.

2025: The client disappears. In January, OpenAI launched Operator, an agent that navigated web sites, crammed varieties, and accomplished purchases autonomously. In Could, Google introduced “Buy for Me” at I/O 2025. In September, On the spot Checkout went stay in ChatGPT.

The friction eliminated: The final one. The human not must be there for the transaction to occur.

Every of those shifts was about the identical factor: eradicating yet another step between wanting and having. Agentic commerce removes the ultimate step: doing it your self.

Checkout Is No Longer A Web page

Right here’s the shift in a single sentence: In conventional commerce, the vendor builds the checkout expertise. In agentic commerce, the agent does.

If you purchase one thing on an internet site in the present day, you work together with the service provider’s checkout web page. They designed the shape, they selected the format, they management the circulate. You fill in your particulars, click on “Purchase,” and the fee processes.

In agentic commerce, the AI agent presents the checkout info inside its personal interface. ChatGPT reveals you the product, the value, the delivery choices, throughout the chat. You verify. The agent handles the remaining. The service provider by no means renders a web page. They obtain an API name.

Stripe’s agentic commerce guide places it instantly: “The components of commerce that was once consumer expertise issues have gotten protocol issues.” As a substitute of optimizing button colours and type layouts, retailers are defining API endpoints and product feeds. Discovery, comparability, and checkout are all dealt with by the agent. The service provider’s job shifts to supplying structured product information and processing the order.

Emily Glassberg Sands, Stripe’s Head of Data and Information Science, framed the broader implications: “Brokers don’t simply change who’s on the checkout. They modify who’s doing the looking out, the deciding, the trusting. All of it.”

I discussed this with Jes Scholz, who ran digital throughout 140+ ecommerce manufacturers at Ringier, on the podcast. Her expertise was clear: Brokers browse in textual content mode, and if they’ll’t parse your web site cleanly, they go away. No second probabilities.

This isn’t theoretical. As of February 2026, a number of agentic commerce implementations are stay. ChatGPT On the spot Checkout is out there to U.S. customers on Free, Plus, and Professional plans. Etsy, Instacart, and Walmart are among the many retailers processing orders by way of it. Shopify’s Agentic Storefronts are energetic by default for eligible retailers, syndicating merchandise to ChatGPT, Google AI Mode, Microsoft Copilot, and Perplexity concurrently. Perplexity launched Instant Buy with PayPal in November 2025, permitting purchases instantly throughout the chat interface with retailers like Wayfair, Abercrombie & Fitch, and 1000’s extra through BigCommerce and Wix.

Each main AI firm is transferring on this path. Anthropic, the corporate behind Claude, has been equally explicit about its commerce plans. In February 2026, Anthropic confirmed it’s constructing options for “agentic commerce, the place Claude acts on a consumer’s behalf to deal with a purchase order or reserving finish to finish,” whereas committing to conserving the expertise ad-free with no sponsored hyperlinks or third-party product placements. Claude already connects to Stripe, PayPal, and Square via MCP integrations. And in June 2025, Anthropic revealed Project Vend, a analysis experiment the place Claude autonomously operated a bodily retail retailer for a month, managing stock, pricing, provider relations, and buyer interactions. The outcomes have been instructive: The agent carried out properly at provider discovery and customer support, however offered objects at a loss and hallucinated fee particulars. A helpful preview of each the potential and the present limitations.

Two open protocols are making this attainable. Each launched inside 4 months of one another.

The Agentic Commerce Protocol

The Agentic Commerce Protocol (ACP) is an open customary co-developed by OpenAI and Stripe, announced Sept. 29, 2025. Licensed beneath Apache 2.0, it defines how AI brokers full purchases on behalf of customers.

ACP makes use of a four-party mannequin: the purchaser (discovers and approves), the AI agent (presents merchandise and handles checkout UI), the service provider (processes the order and fee), and the fee service supplier (handles fee credentials securely). The service provider stays the service provider of report. They course of the fee, deal with success, handle returns. The agent is an middleman, not a market.

The protocol defines four API endpoints:

The duty shift is value spelling out:

The agent handles what the customer sees. The vendor handles what occurs after they click on “Purchase.” ACP will be applied as both a REST API or an MCP server, connecting naturally to the protocol ecosystem lined in Part 3.

Stripe’s Agentic Commerce Suite, launched Dec. 11, 2025, makes ACP adoption sensible. Ahmed Gharib, Stripe’s Product Lead for Agentic Commerce, described it as “a low-code answer enabling companies to promote throughout a number of AI brokers through a single integration.” With out it, connecting to every AI agent individually would take as much as six months of bespoke engineering per platform.

The Suite has three parts: product discovery (sync your catalog and Stripe distributes it to AI brokers), checkout (powered by Stripe’s Checkout Periods API, dealing with taxes and delivery), and funds (utilizing Shared Cost Tokens and Stripe Radar for fraud detection). Retailers join their present product catalog or add on to Stripe, then choose which AI brokers to promote by way of from the Stripe Dashboard.

The ecosystem is rising rapidly. Past OpenAI, Stripe lists Microsoft Copilot, Anthropic, Perplexity, Vercel, and Replit as AI platform companions. On the ecommerce facet, Squarespace, Wix, WooCommerce, BigCommerce, and commercetools have built-in. Salesforce introduced ACP help in October 2025. Shopify’s 1 million+ US retailers are coming quickly.

The Common Commerce Protocol

4 months after ACP launched, a unique coalition unveiled a second customary.

The Universal Commerce Protocol (UCP) was co-developed by Shopify and Google, announced Jan. 11, 2026 on the Nationwide Retail Federation convention in New York. Google CEO Sundar Pichai offered it. The co-developers embody Etsy, Wayfair, Goal, and Walmart. Over 20 corporations endorsed it at launch, together with Mastercard, Visa, Finest Purchase, Dwelling Depot, Macy’s, American Categorical, and Stripe. I broke down UCP and its strategic implications the week it launched on the podcast.

The place ACP is tightly targeted on the checkout circulate, UCP is designed as a full commerce customary masking discovery through post-purchase. Its structure is modeled after TCP/IP, with three layers:

UCP is protocol-agnostic. It helps REST, MCP, A2A, and AP2 (Agent Funds Protocol, Google’s customary for agent-initiated funds). ACP at present helps REST and MCP.

Discovery works by way of a published profile at /.well-known/ucp, much like how A2A brokers publish their capabilities at /.well-known/agent-card.json (lined in Part 3). Each brokers and retailers declare their capabilities, and on every request, the system computes the intersection of what they’ll do collectively. Ashish Gupta, VP/GM of Service provider Buying at Google, described the logic: “The shift to agentic commerce will require a shared language across the ecosystem.”

The 2 protocols mirror totally different strategic positions. ACP, constructed by the corporate operating the AI agent (OpenAI) and the corporate processing the fee (Stripe), is optimized for getting transactions by way of ChatGPT rapidly. UCP, constructed by the corporate internet hosting the retailers (Shopify) and the corporate operating search (Google), is designed for a multi-agent future the place many AI platforms compete for a similar customers.

The excellent news for retailers: These aren’t mutually unique. Shopify retailers can serve each concurrently. The identical merchandise seem in ChatGPT through ACP and in Google AI Mode through UCP. Shopify’s Agentic Storefronts deal with the multi-protocol complexity, syndicating catalog information throughout ChatGPT, Google AI Mode, Microsoft Copilot, and Perplexity from a single admin panel.

Vanessa Lee, Shopify’s VP of Product, framed the corporate’s place: “Agentic commerce has so much potential to redefine shopping and we want to make sure it can scale.”

The Belief Drawback: Funds With out Folks

Each protocols face the identical foundational problem: How do you course of a fee when the individual with the bank card isn’t the one on the checkout?

Conventional commerce treats credential possession as a belief sign. When you have the cardboard quantity, the expiry date, and the CVV, you’re most likely the cardholder. Agentic commerce breaks this assumption. The agent has been approved to behave on the customer’s behalf, but it surely’s not the customer. As Stripe’s Kevin Miller wrote in his October 2025 blog post: “Belief can’t be inferred. It must be explicitly granted, scoped, and enforced in code.”

Javelin Technique & Analysis, cited by Visa, describes this because the shift from “card-not-present” to “person-not-present” transactions. It’s a helpful framing. Card-not-present fraud was the defining problem of ecommerce. Individual-not-present fraud is the defining problem of agentic commerce.

Shared Cost Tokens

Stripe’s answer is the Shared Cost Token (SPT), a brand new fee primitive designed particularly for agent transactions. Right here’s the way it works:

1. The client saves a fee technique with the AI platform (e.g., ChatGPT).
2. When approving a purchase order, the AI platform points an SPT scoped to the particular service provider, capped on the checkout quantity, with a time restrict.
3. The AI platform sends the SPT to the service provider through ACP.
4. The service provider creates a Stripe PaymentIntent utilizing the token.
5. Stripe processes the fee, making use of fraud detection in actual time.

The client’s precise card particulars are by no means shared with the service provider or the agent. Every token is programmable (scoped by service provider, time, and quantity), reusable throughout platforms, and revocable at any time. For present Stripe retailers, enabling SPTs requires “as little as one line of code.”

The Cost Networks Reply

The cardboard networks have launched their very own requirements. Visa launched the Trusted Agent Protocol in October 2025, an open framework constructed on HTTP Message Signatures that helps retailers distinguish professional AI brokers from malicious bots. Developed in collaboration with Cloudflare, it has suggestions from Adyen, Checkout.com, Microsoft, Shopify, Stripe, and Worldpay, amongst others.

Mastercard launched Agent Pay in April 2025, introducing “Agentic Tokens” that construct on its present tokenization infrastructure. Every agent motion makes use of permissions and limits outlined by the buyer. Mastercard CEO Michael Miebach described agent-led funds as a “vital paradigm shift” for the business. U.S. issuers have been enabled in November 2025, with world rollout in early 2026.

PayPal joined the ACP ecosystem on October 28, 2025, enabling PayPal wallets for ChatGPT checkout and constructing an ACP server that connects its world service provider catalog with out requiring particular person service provider integrations.

Google launched its personal fee customary in parallel. The Agent Payments Protocol (AP2), introduced September 2025 with 60+ industry partners, makes use of Verifiable Digital Credentials and a cryptographic Mandate system to create tamper-evident proof of consumer consent at each step of the transaction. AP2 is payment-agnostic, supporting credit score and debit playing cards, real-time financial institution transfers, and even stablecoins through a Coinbase x402 extension. It’s built-in instantly into UCP.

Fraud With out Fingerprints

Conventional fraud detection depends on human behavioral alerts: mouse actions, typing patterns, searching habits, session length. AI brokers have none of those. A professional agent transaction can look indistinguishable from a complicated bot assault.

Stripe addressed this by constructing what they describe as “the world’s first AI foundation model for payments,” a transformer-based mannequin skilled on tens of billions of transactions. The mannequin treats every cost as a token and habits sequences as context, ingesting alerts together with IPs, fee strategies, geography, machine traits, and service provider traits. When SPTs are used, Stripe Radar relays threat alerts together with dispute chance, card testing detection, and stolen card indicators to assist “differentiate between high-intent brokers and low-trust automated bots.”

The assault floor can be novel. Researchers demonstrated in a June 2025 study that ecommerce brokers are vulnerable to visible immediate injection: malicious content material embedded in product listings can hijack agent habits throughout buying duties. All brokers examined have been susceptible. A separate research accepted to IEEE S&P 2026 discovered that 13% of randomly chosen ecommerce web sites had already uncovered their chatbot plugins to oblique immediate injection through third-party content material like product opinions. And a January 2025 paper on authenticated delegation argues that for agentic commerce to operate at scale, the business wants standardized mechanisms to “explicitly delegate authority to brokers, transparently determine these brokers as AI, and implement human-centered decisions round safety and permissions.” SPTs, the Trusted Agent Protocol, and Agent Pay are all early solutions to that problem.

The priority is actual on the buyer facet, too. 88% of consumers surveyed by Javelin are involved that AI shall be used for id fraud, in response to Visa’s evaluation. Constructing belief infrastructure that works for agent transactions is the prerequisite for agentic commerce scaling past early adopters.

→ Learn Extra: Trust In AI Shopping Is Limited As Shoppers Verify On Websites

Who’s Already Promoting to AI

Regardless of the infrastructure nonetheless being constructed, adoption is transferring quick.

AI platforms with commerce capabilities:

Retailers and types on board:

The early adopter record reads like a mall listing. URBN (dad or mum of Anthropologie, Free Folks, and City Outfitters), Etsy, Coach, Kate Spade, Glossier, Vuori, Spanx, SKIMS, Ashley Furnishings, Revolve, and Halara are amongst these onboarding to Stripe’s Agentic Commerce Suite. Walmart and Instacart are stay on ChatGPT. Gymshark, Everlane, and Monos are stay on Google AI Mode through UCP.

Ecommerce platforms enabling it:

Shopify’s 1 million+ U.S. retailers are eligible for ChatGPT integration. BigCommerce, Wix, Squarespace, WooCommerce, and commercetools have built-in with Stripe’s Suite. Salesforce Commerce Cloud introduced ACP help in October 2025, with new Agentforce brokers for service provider, purchaser, and private shopper workflows.

The Market

The market projections fluctuate broadly, which tells you ways early we’re. McKinsey tasks $1 trillion in U.S. retail income orchestrated by brokers by 2030, scaling to $3-5 trillion globally. Gartner predicts 90% of B2B purchases shall be dealt with by AI brokers inside three years, intermediating $15 trillion in spending by 2028. Forrester predicts that by 2026, one-third of retail market tasks shall be abandoned as reply engines steal site visitors.

The buyer facet is extra cautious. A Contentsquare survey of 1,300 U.S. shoppers discovered 30% prepared to let an AI agent full a purchase order on their behalf. A YouGov survey of 1,287 U.S. adults discovered 65% belief AI to match costs, however solely 14% belief it to really place an order. Amongst Gen Z, that quantity rises to twenty%. The hole between “I’ll let AI assist me store” and “I’ll let AI purchase for me” is the place we’re proper now.

However the site visitors is already there. AI-driven site visitors to U.S. retail web sites grew 4,700% year-over-year by mid-2025, in response to Adobe Analytics. Shopify reported that orders attributed to AI searches grew 11x since January 2025. OpenAI estimates roughly 2% of all ChatGPT queries are shopping-related, roughly 50 million buying queries each day throughout a consumer base of 700 million weekly customers.

Educational analysis is beginning to reveal what occurs when brokers do the shopping for. A Columbia Business School and Yale study (August 2025) launched ACES, the primary agentic ecommerce simulator, and examined six frontier fashions, together with Claude and GPT-4. They discovered that AI buying brokers exhibit “selection homogeneity,” concentrating demand on a small variety of merchandise and exhibiting robust place biases in how listings are ranked. The researchers warn of winner-take-all dynamics and the emergence of “AI-Web optimization,” the place sellers optimize listings particularly for agent habits fairly than human preferences. A February 2026 study on personalised product curation discovered that present agentic techniques stay “largely inadequate” for tailor-made product suggestions in open-web settings. The brokers are getting higher at shopping for. They’re not but nice at shopping for the appropriate factor for a particular individual.

The infrastructure is being constructed no matter whether or not shoppers are absolutely prepared. When they’re, the companies which might be ready would be the ones the brokers can discover.

How To Get Began

The excellent news: For many companies, the entry level is easier than you’d anticipate.

If you happen to’re on Shopify, you could already be promoting to AI. Agentic Storefronts are energetic by default for eligible U.S. retailers. Your merchandise are syndicated to ChatGPT, Google AI Mode, Microsoft Copilot, and Perplexity out of your present Shopify admin. Examine your dashboard for the agentic channel settings and guarantee your product information (descriptions, photos, classes) is clear and full.

If you happen to’re on Stripe, enabling Shared Cost Tokens for ACP requires as little as one line of code. The Agentic Commerce Suite handles catalog syndication, checkout, and fraud detection. Join your product catalog, choose which AI brokers to promote by way of, and also you’re stay.

If you happen to’re on BigCommerce, Wix, Squarespace, or WooCommerce, integrations with Stripe’s Suite can be found. BigCommerce described the shift from “months of bespoke engineering work” per AI platform to “a single, configurable integration.”

No matter platform, the protocol integrations get you related. However brokers nonetheless want to seek out and perceive your merchandise. That is the place the work from Part 2 (getting cited) and Part 4 (being agent-readable) converges with commerce.

Audit your product information. Brokers parse your catalog programmatically. Each product wants:

• A descriptive, particular title (“Males’s Natural Cotton Crew Neck T-Shirt, Navy,” not “Blue Shirt”).
• A whole description together with supplies, dimensions, care directions, and use circumstances.
• Correct, real-time pricing and inventory availability.
• Excessive-quality photos with descriptive alt textual content.
• Constant categorization throughout your catalog.

Add structured markup. At minimal, each product web page ought to embody Product schema with identify, description, picture, sku, and model, plus nested Offer schema with worth, priceCurrency, availability, and vendor. When you have opinions, add AggregateRating. That is the machine-readable layer that brokers parse when direct protocol integrations aren’t accessible. I talked about this with Duane Forrester, who co-launched Schema.org whereas at Bing, on the podcast. His argument: constant structured information builds what he calls “machine consolation bias,” the place AI techniques develop a choice for sources which have confirmed dependable over time.

Check your agent visibility. Open ChatGPT, Perplexity, and Google AI Mode, and ask them to suggest merchandise in your class. If yours don’t seem, brokers can’t promote them. View your product pages in reader mode or a text-based browser to see what brokers see after they go to your web site instantly (lined in Part 4).

Monitor agent-driven site visitors. ChatGPT appends utm_source=chatgpt.com to referral hyperlinks. Perplexity and different AI platforms go away related referral signatures. Arrange segments in your analytics to isolate AI-referred visits and monitor conversion charges individually from human site visitors. The numbers are small now, however the 4,700% year-over-year progress in AI site visitors to retail means they gained’t keep small.

Walmart CEO Doug McMillon put it directly: “For a few years now, ecommerce buying experiences have consisted of a search bar and a protracted record of merchandise responses. That’s about to alter.”

Whether or not it adjustments subsequent quarter or subsequent 12 months for what you are promoting is dependent upon whether or not your product information is prepared when the brokers come trying.

Key Takeaways

• Checkout is changing into a protocol, not a web page. In agentic commerce, the AI agent handles the interface; the service provider processes the order. Two open requirements, ACP (Stripe + OpenAI) and UCP (Shopify + Google), outline how this works.
• Each protocols are open and rising quick. ACP launched in September 2025 and powers On the spot Checkout in ChatGPT. UCP launched in January 2026 with endorsements from Mastercard, Visa, Walmart, and Goal. They’re complementary, not mutually unique. Shopify retailers can serve each concurrently.
• Shared Cost Tokens remedy the “person-not-present” drawback. When the customer isn’t on the checkout, conventional belief alerts break down. SPTs are programmable, scoped, time-limited, and revocable, letting brokers provoke funds with out ever seeing the customer’s card particulars.
• Cost networks are constructing their very own requirements. Visa’s Trusted Agent Protocol and Mastercard’s Agent Pay present authentication and fraud frameworks particular to agent transactions. PayPal joined the ACP ecosystem. The funds infrastructure for agentic commerce is taking form throughout the business.
• Main manufacturers are already stay. Etsy, Walmart, Instacart, Glossier, SKIMS, Coach, and dozens extra are promoting by way of AI brokers in the present day. Ecommerce platforms, together with Shopify, BigCommerce, Wix, Squarespace, and WooCommerce, have integrations accessible.
• Shopper belief is lagging behind infrastructure. Solely 14% of shoppers at present belief AI to position orders on their behalf. However AI-driven site visitors to retail grew 4,700% in a 12 months. The infrastructure is being constructed for the adoption curve that follows.

That is the ultimate article in a five-part sequence on the agentic net. Part 1 framed the shift from Web optimization to AAIO. Part 2 lined the right way to get cited by AI. Part 3 mapped the protocols. Part 4 defined how brokers understand your web site. This text lined the place all of it leads: transactions.

The thread connecting all 5 components is simple. Structured information helps AI discover you. Clear content material helps AI cite you. Accessible HTML helps AI navigate you. Structured commerce protocols assist AI purchase from you. It’s the identical precept at each layer: Make what you are promoting machine-readable, and the machines will do enterprise with you.

Kevin Miller, Stripe’s Head of Funds, captured the moment: “Stripe spent the final 15 years optimizing commerce for human customers. Now, we’re beginning to do the identical with brokers.”

The brokers are already buying. The query is whether or not they can discover your retailer.

Extra Sources:

This publish was initially revealed on No Hacks.

Featured Picture: showcake/Shutterstock