OpenAI Agentic Commerce Launch: Instant Checkout Explained (2025)

Reading Time: 11 minutes

---

From Chat to Checkout – The Commerce Landscape Just Shifted

Over 700 million people turn to ChatGPT each week for everyday tasks, from researching products to comparing options. Until now, that’s where the journey ended. Users would click out to retailers websites, re-enter their searches, and complete purchases through traditional checkout flows.

That journey is about to disappear.

Starting today (29th September 2025), US-based ChatGPT users (Plus, Pro, and Free tiers) have the ability to purchase directly from Etsy sellers without leaving the chat interface. Over a million Shopify merchants – including major brands like Glossier, SKIMS, Spanx and Vuori are coming online soon. Currently limited to single item purchases, the instant checkout roadmap includes multi-item carts and expansion to new regions and merchants.

The thing is however, OpenAI is not simply launching an instant checkout in their platform – their open-sourcing the building blocks behind the tech to allow merchants, and developers (like flowio) to build their own agentic checkout systems.

What Is The Agentic Commerce Protocol?

Co-developed with Stripe and leading merchant partners, the Agentic Commerce Protocol is an open standard that provides the language for AI agents and businesses to work together to complete purchases for users.

Think of it as the missing infrastructure layer between conversational AI and commerce systems. Previously, connecting an AI agent to payment processing, inventory management, and fulfilment required bespoke integration work for each merchant.

The Agentic Commerce Protocol standardises this, making it possible for any merchant to become “agent-ready” without rebuilding their backend systems.

The protocol has been designed with three core principles:

Platform Agnostic: It works across different AI platforms, payment processors, and business types. This isn’t an OpenAI only standard. Any AI agent can be implemented with it.

Rapid Integration: Merchants using Stripe can enable agentic payments in as little as one line of code. Those using other payment processors can still participate through Stripe’s Shared Payment Token API or by adopting the Delegated Payments Spec.

Merchant Control: Perhaps most importantly, merchants remain the merchant of record throughout the entire purchase journey. They control fulfilment, returns, customer support, and communication. ChatGPT acts purely as the user’s agent – passing information between buyer and seller, much like a digital personal shopper.

How Instant Checkout Actually Works

The user experience is pretty simple. Someone asks ChatGPT a shopping question “best running shoes under £100” or “gifts for a ceramics lover”.

ChatGPT returns relevant products from across the web. These results are organic and unsponsored, ranked purely on relevance.

The user then has the ability to ask follow up questions about the products, compare the price and features against similar products, and ChatGPT surfaces price trends – e.g. “Lowest price in 6 weeks”.

If a product supports Instant Checkout, users can tap “Buy”, confirm their order details, and complete the purchase without leaving chat. Existing subscribers can use their card on file or choose express payment options.

Behind the scenes, ChatGPT sends the necessary order details to the merchant’s backend using the Agentic Commerce Protocol. The merchant accepts or declines the order, processes payment through their existing provider, and handles fulfilment exactly as they do today.

Merchants pay a small fee on completed purchases.

The service is free for users, doesn’t affect pricing, and critically, doesn’t influence product rankings.

Instant Checkout items receive no preferential treatment in search results. When ranking multiple merchants selling the same product, ChatGPT considers availability, price, quality, whether a merchant is the primary seller, and whether Instant Checkout is enabled, purely to optimise user experience.

On top of this, OpenAI has released new Product Feed Specification guidance which, if you’re familiar with Google Merchant Center and running Google Shopping Ads will be similar – yet quite a few differences which we cover later in this article.

Behind The Infrastructure

It comes as no surprise that OpenAI has decided to partner with Shopify, Etsy and Stripe for the launch into agentic commerce. Stripe is by far and wide one of the most used payment systems when it comes to building with AI, and provides an easy integration when building conversational agents.

Stripe – The Backbone Of Agentic Commerce

“Stripe is building the economic infrastructure for AI. That means re-architecting today’s commerce systems and creating new AI-powered experiences for billions of people. We’re proud to power Instant Checkout in ChatGPT and co-develop the Agentic Commerce Protocol to help businesses and AI platforms build the future of commerce.” – Will Gaybrick, President, Technology and Business, Stripe

The partnership goes deeper than just payment processing. Stripe co-developed the Agentic Commerce Protocol itself, bringing decades of experience in secure, compliant payment infrastructure to the table. For merchants already using Stripe, this means enabling agentic payments can be as simple as a single line of code. For those using alternative processors, Stripe has built the Shared Payment Token API specifically to bridge this gap, allowing merchants to participate without switching their entire payment stack.

This is critical for adoption. By not forcing merchants to rip out existing systems, Stripe has removed one of the biggest barriers to integration.

Shopify – The Gateway to Commerce

Shopify is the platform of choice for millions of retailers globally, powering everyone from solo entrepreneurs to billion-pound brands. With over a million merchants coming to Instant Checkout, Shopify represents the primary on-ramp for mainstream retail adoption.

What makes Shopify particularly well-suited for this is their app ecosystem and API architecture. The platform is already designed for third-party integrations, meaning adding Agentic Commerce Protocol support can be rolled out across their entire merchant base relatively quickly.

For many Shopify merchants, this won’t require custom development – it will be a toggle in their admin panel.

The inclusion of high-profile Shopify brands like Glossier, SKIMS, Spanx, and Vuori in the initial launch signals something important: this isn’t just for small merchants. Enterprise-scale retailers are betting on agentic commerce from day one.

Etsy: The Long-Tail Content Play

Etsy’s inclusion is equally strategic but for different reasons. While Shopify brings scale and brand recognition, Etsy brings product diversity and long-tail inventory. When someone asks ChatGPT for “handmade ceramic bowls” or “custom leather journals”, Etsy’s marketplace of unique, artisan products becomes invaluable.

This is where AI-driven discovery particularly shines. Traditional search on Etsy requires users to know what they’re looking for and how to describe it. AI agents can understand intent, ask clarifying questions, and surface relevant products from millions of listings. For Etsy sellers, this could represent a fundamental shift in discoverability, moving from SEO-optimised product titles to semantically rich descriptions that AI agents can understand and recommend.

The combination of Shopify (mainstream retail), Etsy (unique/handmade), and Stripe (universal payment infrastructure) gives OpenAI coverage across most e-commerce use cases while keeping technical integration manageable.

---

The Product Feed Spec: What’s Different From Google Shopping

If you’ve optimised Google Shopping feeds like I have, OpenAI’s Product Feed Specification will feel familiar – until it doesn’t. The structure borrows heavily from Google Merchant Center, but the opportunities are entirely new.

The familiar bits: Standard formats (TSV, CSV, XML, JSON), core required fields (id, title, description, price, availability), and HTTPS delivery. If your Google Shopping feed works, you’re 70% there.

The new bits that matter:

1. Granular Control Flags

Two new required fields: enable_search and enable_checkout. You can make products discoverable in ChatGPT without enabling direct purchase. Want brand visibility but prefer handling transactions on your site? Set enable_search: true and enable_checkout: false. Google Shopping doesn’t offer this middle ground.

2. Real Inventory Numbers

Google Shopping lets you be vague about availability. OpenAI requires inventory_quantity – an actual integer stock value. ChatGPT can now say “3 left in stock” or “only 1 available”, creating urgency and transparency. But your inventory systems must be accurate in real-time. Pushing feeds every 15 minutes while inventory updates hourly? You might need to have a think around this for high throughput items.

3. Performance Signals

popularity_score and return_rate are recommended fields that don’t exist in Google Shopping. A product with a 4.8 popularity score and 2% return rate will likely outrank identical products without these signals. If your e-commerce platform tracks this data (most do), surfacing it could materially impact visibility.

4. Raw Review Content

The feed accepts raw_review_data and q_and_a – full-text content, not just ratings. If someone asks “are these shoes good for wide feet?”, ChatGPT can surface your product because a reviewer mentioned it, even if that’s not in your description. Google Shopping can’t do this. It ranks on structured attributes. ChatGPT ranks on semantic understanding of everything.

Detailed, specific reviews now matter more than star ratings.

5. Custom Variant Fields

Three custom variant options (custom_variant1_category through custom_variant3_category) let you define what makes your products unique. Selling furniture? Use Wood_Type: Oak/Mahogany/Walnut and Finish: Matte/Glossy. Selling tech? Use Storage_Capacity: 256GB/512GB and Warranty_Level: Standard/Premium.

Google Shopping forces you into predefined attributes. OpenAI doesn’t.

6. Product Relationships

related_product_id and relationship_type fields let you explicitly define that products are part_of_set, often_bought_with, or substitute for each other. ChatGPT can then make intelligent recommendations: “Would you like the matching belt?” or “That’s out of stock, but this alternative has better reviews.”

7. Geo-Specific Everything

geo_price and geo_availability allow region-specific overrides in a single feed. Different prices by state? Variable stock by region? Specify it granularly. Google Shopping requires separate feeds per country. OpenAI consolidates regional variations.

8. Video Thumbnails & 3D Models

Whilst image links are required, GPT’s product feed spec lists model_3d_link allowing GLB/GLTF 3D model links. I think this opens up a door to visual possibilities to display products, technical or otherwise. I imagine these will be interactive within the chat interface. video_link also provides the possibility that we can serve video thumbnails rather than static images which again improves the visual elements over Google Shopping ads.

The Optimisation Shift

This is the fundamental difference:

Google Shopping rewards: Keyword-stuffed titles, competitive bids, click-through rates ChatGPT rewards: Semantic clarity, comprehensive attributes, performance signals, contextual understanding

Your Google Shopping title: “Men’s Running Shoes Blue Size 10 Waterproof Trail”

Your ChatGPT title: “Men’s Trail Running Shoes – Blue” Your ChatGPT description: “Waterproof trail running shoe designed for challenging terrain, featuring enhanced grip and cushioned midsole for long-distance comfort.”

The AI agent reads semantically. It doesn’t need keyword stuffing. It will need context.

The 15-Minute Reality

OpenAI accepts feed updates every 15 minutes. Google Shopping updates daily (or hourly depending on your setup).

The opportunity: Dynamic pricing and real-time inventory reflected almost immediately. Drop your price at 2pm, ChatGPT users see it by 2:15pm.

The pressure: Your systems must generate accurate feeds every 15 minutes. If you’re exporting from your e-commerce platform once daily via cron job, you’ll need to rebuild your data pipeline.

For technically sophisticated retailers with real-time systems, this is an advantage. For smaller merchants managing spreadsheets, it’s a significant hurdle.

However, even with the most basic of product feeds – there is always the possibility to improve data pipeline, and automate feed optimisation with a partner that has experience in product feed management and AI automation such as flowio.

---

Why This Matters For UK Retailers

While the initial rollout is US only, UK retailers should be paying close attention.

The Search Behaviour Shift Is Already Happening

Consumers are increasingly using AI chatbots as their primary search and discovery tool. This isn’t speculative, 700 million people per week are already doing it. When commerce capabilities are baked directly into these interfaces, the entire purchase funnel collapses. Discovery, evaluation, and purchase happen in the same conversation.

Traditional e-commerce depends on driving traffic to your website, where you control the user experience, collect data, and optimise conversion. Agentic commerce inverts this. The AI agent controls the interface, you’re competing on product relevance, availability, and price transparency, not landing page design or SEO.

Data Sovereignty Becomes Even More Critical

As we’ve written previously about the importance of data sovereignty for UK businesses, the rise of agentic commerce adds another dimension to this conversation. When AI agents are facilitating purchases, who owns the customer data?

What happens to browsing behaviour, purchase history, and preference signals?

The Agentic Commerce Protocol explicitly keeps merchants as the merchant of record, which helps. But retailers need to understand what data they’re sharing with AI platforms, what they’re receiving back, and how that impacts their ability to build customer relationships and personalised experiences.

First-Mover Advantage Is Real

When this launches in the UK, there will be a window where adoption is low and visibility is high. Early adopters will benefit from being among the limited set of merchants available through AI-driven purchase flows. As more merchants integrate, competition increases and differentiation becomes harder.

Consider the Shopify ecosystem. Over a million merchants are coming online soon. Those who integrate early will have months of data, optimised product descriptions for AI discovery, and refined processes before the majority of their competitors even understand what’s happening.

It’s Not Just About ChatGPT

The protocol is open-sourced. This means any AI agent can implement it. Google’s Gemini, Anthropic’s Claude, Microsoft’s Copilot, any conversational AI platform could adopt this standard and instantly unlock commerce capabilities. It may even be possible to build agentic checkout systems through WhatsApp conversational chatbots.

It’s about preparing for a fundamental shift in how AI agents interact with commerce infrastructure.

What Should UK Retailers Do Now?

1. Understand Your Payment Infrastructure

If you’re already using Stripe, integration could be as simple as enabling a feature flag. If you’re using another processor, you’ll need to evaluate whether to use Stripe’s Shared Payment Token API or implement the Delegated Payments Spec directly.

2. Audit Your Product Data

AI agents rank products based on relevance. That means your product descriptions, categorisation, attributes, and metadata need to be optimised not for human readers, but for semantic understanding. Vague descriptions and marketing fluff won’t cut it. Clear, specific, attribute-rich product information will.

3. Monitor The Documentation

OpenAI has made the protocol documentation and integration guides publicly available. Even if you’re not ready to integrate immediately, understanding the technical requirements now means you can move quickly when UK availability launches.

4. Consider Your Inventory Systems

Agentic commerce will put new demands on real-time inventory accuracy. If an AI agent shows a product as available, but you’re out of stock, you’ve created a poor user experience and potentially lost the sale. Systems that work for traditional e-commerce may not be responsive enough for agent-driven purchases.

5. Think Beyond Immediate Sales

The more interesting question isn’t “how do we sell through ChatGPT”, it’s “how does our business model change when AI agents mediate the entire customer relationship?” Do you optimise for agent-driven discovery? Do you build your own AI agents? Do you focus on products where human expertise still matters?

Getting Started

Building with the Agentic Commerce Protocol is open to all. Instant Checkout in ChatGPT is currently available to approved partners. To make your products available for Instant Checkout through ChatGPT, please do the following:

1. Apply to participate in Instant Checkout.
2. Share your product feed according to our Product Feed Spec in order to provide ChatGPT with accurate, up-to-date information about your products.
3. Build your Agentic Checkout API according to the Agentic Checkout Spec. This involves: a. Implementing the required REST endpoints b. Implementing webhooks to notify OpenAI of order events c. Returning rich checkout state on every response
4. Build your payments integration. Use a trusted payment service provider (PSP) that is compliant with the Delegated Payment Spec in order to securely transmit and charge payment credentials. Stripe’s Shared Payment Token is the first Delegated Payment Spec-compatible implementation with more PSPs coming soon. If you’re a PSP or a PCI DSS level 1 merchant with your own vault, learn how to build a direct integration with OpenAI.
5. Certify with OpenAI and move to production. To ensure products, payments and orders are all working correctly, work with OpenAI to pass conformance checks and receive production access.

OpenAI plans to onboard new partners on a rolling basis, beginning in the U.S. If you’re an Etsy or Shopify merchant, you do not need to apply or build an integration as you are already eligible.

The Bigger Picture

Agentic commerce represents the next phase in how AI systems interact with the world on our behalf. We’ve moved from AI that answers questions, to AI that takes actions, to AI that completes transactions.

For retailers, this is both an opportunity and a challenge. The opportunity is reaching hundreds of millions of people through a frictionless purchase experience. The challenge is maintaining differentiation and customer relationships when an AI agent sits between you and your buyers.

The retailers who will thrive in this environment aren’t necessarily those with the best websites or the largest marketing budgets. They’re the ones who understand how to make their products discoverable, their inventory systems reliable, and their integration seamless for AI agents acting on behalf of human shoppers.

The protocol is open-sourced. The documentation is available. UK expansion is inevitable, and it opens a massive opportunity for early adopt UK retailers.

Will you be ready for agentic commerce when it arrives?