OpenAI Scales Back Instant Checkout — Structural Challenges of Agentic Commerce Exposed

Instant Checkout in ChatGPT Shifts to Retailer Apps

OpenAI Scales Back On Instant Checkout Feature: What Does This Mean For Agentic Commerce?

OpenAI has pulled back from letting shoppers pay in ChatGPT.

On March 17, 2026, following a report by The Information, OpenAI announced its decision to scale back the rollout of its native payment feature "Instant Checkout" within ChatGPT. Going forward, when users find products on ChatGPT, instead of completing payment within the chat interface, they will be redirected to partner retailer apps such as Instacart, Target, and Expedia to make their purchases. An OpenAI spokesperson confirmed to Search Engine Land that "Instant Checkout is moving to apps, with purchases happening more seamlessly within connected services."

Background and Industry Trends

In September 2025, OpenAI launched Instant Checkout alongside the "Agentic Commerce Protocol (ACP)" co-developed with Stripe, drawing attention as a pioneer in agentic commerce. The rollout began with Etsy sellers in the US, with plans to expand to over one million Shopify merchants. Bernstein analysts hailed the move as "the beginning of Agentic One."

However, reality proved harsh. According to Shopify President Harley Finkelstein, only "about 12 stores" among Shopify merchants were actively using AI tools. Compared to Shopify's total merchant base of several million, this was effectively zero. OpenAI discovered that while users were eager to research products on ChatGPT, they would not complete payments there. Moreover, reports indicated that as of February 2026, OpenAI had not even established a system for collecting and remitting state taxes, underscoring how minimal transaction volumes had been.

"The Demo-to-Production Gap" — Five Structural Constraints

This strategic pivot reflects not just an OpenAI-specific issue, but structural challenges facing agentic commerce as a whole.

Chris Jones, Managing Director at PSE Consulting, told TechRound that "this should not be seen as a failure of agentic commerce." He cited five structural constraints identified in an Adyen research report surveying over 200 enterprise merchants.

• Protocol fragmentation — Multiple standards such as ACP and UCP coexist without unification
• Unstructured product data — Data not designed to be machine-readable
• Legacy enterprise systems — Inventory management, order processing, and logistics running on outdated stacks
• Trust and liability frameworks — Unclear accountability when AI agents execute payments
• Large-scale onboarding — Immature mechanisms for connecting thousands of merchants simultaneously

Jones points out that "what works in a demo breaks easily in production." A checkout experience spanning thousands of retailers, multiple currencies, and various national tax systems cannot be contained within a single chatbot.

The "Trust" Problem That Payment Alone Cannot Solve

Lee McKenzie, Director of Online Visibility at Semrush, identified two forces hindering the adoption of agentic commerce in comments to Search Engine Land.

The first is an infrastructure challenge. Real-time catalog normalization at the scale of tens of millions of SKUs is a capability Google has spent a decade building through "Merchant Center" — it cannot be replicated overnight. The second is consumer trust. Most users prefer payment flows they are already familiar with, such as Apple Pay, Google Wallet, and Amazon's one-click purchasing. The psychological barrier to entrusting credit card information to an unfamiliar AI chat interface remains high.

Adyen's research shows that while more than half of US consumers expressed willingness to delegate purchasing to AI, 27% of retailers cited "loss of direct customer relationships" as a primary concern. The volume of agentic commerce transactions remains at "immaterial levels."

Impact and Action Items for E-Commerce Businesses

OpenAI's strategic pivot sends a clear signal to e-commerce businesses.

Go all-in on "discovery." ChatGPT continues to function as a venue for product search, comparison, and recommendation. According to Style Arcade data, AI agent-driven online sales reached $14.2 billion globally and $3 billion in the US alone during Black Friday 2025. The power of AI-driven "discovery" has been proven.

Maintain existing trusted flows for "payment." There is no need to rush into supporting native AI checkout. Now that OpenAI itself has switched to a retailer app redirect model, businesses should focus on optimizing the purchasing experience within their own apps and websites.

Structuring product data is urgent. To enable AI agents to accurately understand products, making information such as sizing, inventory, and return policies machine-readable should be the top priority. The fashion industry averages 18% excess inventory and 22% demand mismatch — these are problems of inventory management and data accuracy, not AI checkout buttons.

Conclusion

OpenAI's scaling back of Instant Checkout is not the "end" of agentic commerce, but a "reality check." As Jones states, "The ChatGPT and Adyen developments are not a step backward — they show that the infrastructure to make autonomous shopping safe, reliable, and seamless is still being built."

OpenAI will continue developing ACP in partnership with Stripe while shifting toward supporting app-based transactions. ChatGPT remains "where you decide what to buy," but it is no longer "where you buy." For e-commerce businesses, the next key developments to watch are the standardization race between Google's UCP and ACP, and how far payment infrastructure providers (Stripe, Adyen, Mastercard) will go in implementing agent support.