- The Agentic Commerce Protocol (ACP) from OpenAI and Stripe connects ChatGPT to merchants' systems. Since March 2026, its main use has been product search in ChatGPT.
- Instant Checkout, which let users buy directly in the chat, launched in the US in September 2025. In March 2026, OpenAI changed course: merchants are back to using their own checkout.
- Shopify merchants are represented in ChatGPT through Shopify Catalog with no effort on their part. Purchases happen in the merchant's store, so the merchant remains the seller and keeps the customer relationship.
- Whether ChatGPT recommends a product depends above all on complete, up-to-date product data. Google is pushing its own standard, the Universal Commerce Protocol (UCP).
With the Agentic Commerce Protocol (ACP) and the Shopify integration, part of the shopping journey is moving into AI. Products are no longer found only through a search bar, but in conversation with ChatGPT. The AI asks about budget, occasion and requirements, then suggests suitable products.
A lot has changed since we first published this article in November 2025. OpenAI has scaled back buying directly in the chat to focus on product search, and Google has launched its own standard, the Universal Commerce Protocol. This article covers where things stand as of October 2026, what that means for Shopify merchants and where the levers are for profitable revenue growth.
Opportunities and risks have shifted since the launch
- The opportunity: buyers close to a decision. ChatGPT helps users find and compare products. According to OpenAI, this brings merchants buyers with higher purchase intent who are close to making a decision.
- The risk: a new dependence on a platform. OpenAI decides which products ChatGPT shows. Merchants can influence this only through the quality of their data.
- The consequence: product data becomes the biggest lever. To be recommended in ChatGPT, you need complete, accurate and up-to-date product data that an AI can understand and compare.
ACP connects AI agents with your merchant systems
The Agentic Commerce Protocol is an open standard developed by OpenAI together with Stripe and released as open source in September 2025. It defines how buyers, their AI agent and the business interact to complete a purchase.
This turns an AI assistant that provides recommendations and links into an agent that can handle parts of the purchase, from research to ordering. Technically, ACP is not a new store platform that replaces Shopify. It's a shared interface between the AI front end and merchants' existing infrastructure: store system, payment provider and inventory management. Instead of building a separate integration for every store, everyone speaks the same language.
Stripe provides the payment layer. With a Shared Payment Token, ChatGPT can initiate a payment on the user's behalf without passing on their full payment details. The token is valid only for a specific amount and a specific merchant. Merchants using other payment providers can participate via the Delegated Payments specification.
Since March 2026, OpenAI has expanded ACP into the connection layer for all product search. Merchants use it to share product feeds and promotions, including through third parties such as Salesforce and Stripe. According to OpenAI, early adopters of ACP for product search include Target, Sephora, Nordstrom, Lowe’s, Best Buy, The Home Depot and Wayfair.
Instant Checkout revealed the limits of buying in the chat
On September 29, 2025, OpenAI launched Instant Checkout: users in the US could buy from Etsy sellers directly in the chat without leaving ChatGPT. More than one million Shopify merchants were set to follow, including Glossier, SKIMS, Spanx and Vuori. At launch, only single-item purchases were supported. At the time, OpenAI cited more than 700 million weekly users.
How it worked for users:
- Query: The user asks, for example, for the best running shoes under 100 dollars.
- Recommendation: ChatGPT shows relevant products. According to OpenAI, results are organic, not sponsored, and ranked by relevance.
- Purchase: A “Buy” button appears for supported products. The user confirms order, shipping and payment in the chat.
- Fulfillment: Order, payment and shipping run through the merchant's existing systems.
Merchants paid OpenAI a fee on completed purchases. Alex Sperber's assessment below dates from the time when purchases were still meant to happen entirely in the chat:
The technical integration may be seamless thanks to Shopify's ‘zero integration’ approach, but the strategic implications are brutal. This is the classic platform dilemma 2.0. Merchants are trading their most valuable currency, the direct customer relationship and the email opt-in, for short-term sales volume.
Alex Sperber, Partner at UnitedAds
In March 2026, OpenAI changed course, saying the first version of Instant Checkout had not offered the flexibility it was aiming for. Merchants are therefore back to using their own checkout, and OpenAI is focusing on product search. Merchants who want deeper integration can build their own app in ChatGPT; Walmart was the first to do so, with customer accounts, a loyalty program and its own payment.
Shopify merchants are represented in ChatGPT with no effort on their part
For Shopify merchants, the key news is that they don't have to integrate anything. According to OpenAI, their product data already flows into ChatGPT through Shopify Catalog, with no extra work for individual merchants. Shopify handles the connection at the platform level and aggregates its merchants' catalogs.
How it has worked since March 2026:
- Product data: Shopify supplies product information, prices and availability to ChatGPT in aggregated form.
- Product search: A user discovers a product in conversation with ChatGPT, with images, prices, reviews and comparisons.
- Purchase: The user completes the purchase in the merchant's online store, via an in-app browser in ChatGPT.
- Fulfillment: The order lands in the Shopify admin as usual.
The merchant remains the merchant of record: they sell, ship, and handle returns and customer service. Because the purchase takes place in their own store, the merchant keeps control of the brand experience and the customer relationship.
Merchants who don't sell on Shopify need their own route into ChatGPT: a product feed that follows the ACP specification, submitted directly or via providers such as Salesforce and Stripe. We explain how this works in our article on ChatGPT's shopping feature.
Your product data determines whether ChatGPT recommends your products
There's no second results page in AI search. ChatGPT names just a few products, and which ones depends heavily on the data the merchant provides. According to OpenAI, required fields ensure accurate prices and availability, while additional information such as reviews improves ranking and relevance.
For Google search, by contrast, Google says that optimizing for AI search is still SEO, so classic SEO remains important. ChatGPT adds a second discipline, which we call Generative Engine Optimization (GEO): preparing product catalogs and content so AI systems can understand, compare and recommend them.
How Shopify merchants can prepare their product data for AI
- Accurate, complete descriptions: The AI needs materials, dimensions, use cases and concrete features. Generic marketing copy doesn't help. Good descriptions reflect real customer intent, such as “waterproof jacket for fall hikes”.
- Consistent tags with context: A consistent tagging system for occasion (“wedding”, “office”), style (“minimalist”, “vintage”) and use (“travel”, “home office”) helps the AI match conversational queries to the right products.
- Clean master data: accurate GTINs and SKUs, a clear variant structure for size and color, reviews, questions and answers, and clear return policies.
- Up-to-date prices and stock levels: If the AI shows the wrong price or stock level, you lose both the sale and the buyer's trust. Update price and availability as often as possible.
Accuracy matters twice over: the AI fills in whatever the data leaves open, which can produce inaccurate descriptions that mislead buyers.
Merchants who already manage their product data in a PIM system have an advantage. Detailed attributes, consistent tags and clean master data are core tasks of product data management. Merchants still managing their data in spreadsheets will struggle to deliver the feeds an AI needs.
| Area | Classic search | AI search in ChatGPT |
|---|---|---|
| Goal | Ranking on the first results page | Recommendation in the answer |
| Search queries | short keywords, such as “red women's shoes” | conversational intent, such as “comfortable red pumps for a wedding” |
| Most important data source | Website content, collected by the crawler | Product feed, delivered directly to the AI |
| Content | readable descriptions for people and crawlers | detailed attributes: material, dimensions, use, benefits |
| Technology | structured data, page speed, backlinks | accurate GTINs and SKUs, tags with context, reviews |
| Freshness | regular crawl | feed updates at least daily |
Weighing reach against control, cost and the customer relationship
The return to merchants' own checkout has defused the biggest risks of the first phase. Some trade-offs remain.
The AI controls how your brand is presented
Product search takes place in ChatGPT, not in your store. The AI decides, based on your data, how your products are described, compared and positioned. Your store's design, navigation and curated content only come into play once the user clicks through to you. Upselling and cross-selling are possible only in the store.
The customer relationship stays with you
With the original Instant Checkout, merchants received only the data needed to fulfill the order. Now that purchases happen in the merchant's store again, customer data, email marketing consent and customer accounts stay with the merchant, as with any other store purchase. This is critical for customer lifetime value and one of the most important reasons not to give up the customer relationship.
Factor the costs into every margin calculation
For Instant Checkout purchases, merchants paid OpenAI a fee on top of store system and payment processing fees. Check whether, and which, fees will apply to ChatGPT purchases with every integration. Every channel has to be judged by its contribution to profit, not by revenue.
How results are ranked remains opaque
OpenAI stresses that product results are organic and not sponsored. At the launch of Instant Checkout, however, OpenAI named ranking factors for cases where several merchants sell the same product: availability, price, quality, whether a merchant is the primary seller and whether Instant Checkout is enabled. How ranking works in detail today is not public. Merchants should therefore regularly check for themselves how their products appear in ChatGPT.
| Factor | Opportunity | Risk |
|---|---|---|
| Reach | Access to buyers who begin their product search in ChatGPT | a new dependence on a platform |
| Conversion | Buyers arrive at the store with clear intent | Side-by-side comparison with competitors in the answer |
| Margin | a new channel with no cost per click | check for potential per-sale fees |
| Customer relationship | Purchase in the merchant's own store; data stays with the merchant | for apps or in-chat checkout, depends on the terms |
| Brand | Brand visible in product cards | AI controls description and positioning |
| Visibility | organic results without an ad budget | Ranking logic not transparent in detail |
Several protocols are competing to power agentic commerce
Even when Instant Checkout launched, it was clear that product search in the chat cuts into Google's core business:
What we're seeing here is a direct attack on Google's business model. When the ‘funnel collapses’ and product search moves from the search bar into the chat, Google loses its most lucrative ad slots.
Alex Sperber, Partner at UnitedAds
Google is now pushing its own standard as well. ACP is just one of several. Multiple players are building the foundations for agentic commerce, and merchants should keep a close eye on how it develops.
- ACP (OpenAI and Stripe): pragmatic, built around existing store and payment infrastructure, and since March 2026 used mainly for product search in ChatGPT.
- UCP (Google): Google co-developed the Universal Commerce Protocol. According to Google in May 2026, buyers will soon be able to purchase directly via Google Pay in AI Mode and the Gemini app, or click through to the merchant's store; either way, the merchant remains the merchant of record. Brands named include Nike, Sephora, Target, Walmart and Wayfair, plus Shopify brands Fenty and Steve Madden. After the US, Canada and Australia are next, followed later by the UK; Google has not yet mentioned Germany. Learn more in our article on the Universal Commerce Protocol.
- Payment protocols for agents: Standards are also emerging that let agents authorize payments on users' behalf in a verifiable way, including Google's Agent Payments Protocol (AP2).
- MCP (Model Context Protocol): an open standard that lets AI applications retrieve data from business systems, such as a product catalog or CRM. It complements the transaction protocols: MCP gives the AI the knowledge, while ACP and UCP enable the purchase.
Merchants should prepare for fragmentation. Over the next 24 months, we'll see a situation where you need different optimizations for the Apple/Siri ecosystem, the Google/Gemini ecosystem and the OpenAI ecosystem. The complexity of competing protocols (ACP vs. AP2) will push the dream of ‘set up once, sell everywhere’ far into the future for now.
Alex Sperber, Partner at UnitedAds
Connecting to all these standards would be almost impossible for individual merchants. This is where platforms like Shopify prove their value: they handle the connections to the various protocols and take the technical complexity off their merchants' plates.
Conclusion: Merchants who prepare their product data now will be the ones AI search recommends
Agentic commerce is changing where buying decisions are made. The first phase showed that completing the entire purchase in the chat doesn't work yet. Product search in ChatGPT, however, is growing, and Google is entering the market as a second major player. For merchants, this means the decision increasingly happens in the AI, and the purchase in their own store.
Prepare your product data so that AI systems can clearly match offers, variants and availability. That includes complete attributes, clean master data, current prices and stock levels, and a PIM system as the foundation. Companies that manage their data in silos are mentioned less often in AI answers. You can see how we support retailers with this on our ChatGPT Shopping page.
