The first credible test of AI-native checkout has returned an unambiguous result: retailers will embrace AI as a discovery engine, but not as a point of sale. OpenAI’s pivot to its Agentic Commerce Protocol signals a structural reset – with direct implications for sporting goods brands.
AI commerce has had its first credible industry test – and the outcome is instructive. OpenAI’s Instant Checkout feature, which allowed users to complete purchases directly within ChatGPT without navigating away from the platform, was quietly wound down less than six months after its launch.In a blog post dated March 24, the company confirmed it had ended the program, announcing a pivot toward richer discovery experiences built on the ACP (Agentic Commerce Protocol). The conversion data and retailer feedback that preceded that decision explain both the failure and the direction being taken next.
Conversion rates delivered a clear verdict
The numbers that emerged from the Instant Checkout experiment were unambiguous. Daniel Danker, Executive Vice President of AI acceleration, product and design at Walmart, told Wired that sales through the feature were disappointing – with in-chatbot conversion running three times lower than for products where users clicked through to Walmart’s own site. The friction point was structural: shoppers researching multi-item purchases were reluctant to complete a fragmented transaction that bypassed their existing cart, removed promotional options and created the prospect of multiple separate deliveries.
“They generally don’t want to split the checkout experience,” Danker said, noting that customers with items already in their Walmart cart did not want to purchase a single product through the chatbot while the rest of their basket remained unaffected.
Etsy was among the first merchants to activate Instant Checkout, enabling the feature in September 2025 with millions of products available through the platform. A company spokesperson told Modern Retail that the feature had not delivered meaningful sales volume, though ChatGPT had proven useful as a discovery and referral channel, directing traffic back to Etsy sellers’ product pages. The end of Instant Checkout would not prompt a significant strategy change, the spokesperson added; Etsy is now developing a dedicated ChatGPT app designed to surface its sellers and their original products in a format that gives it greater control over the user experience.
Missing infrastructure made merchant adoption impractical
Feedback from retail executives pointed consistently to capability gaps that were never resolved. A C-suite executive at a large specialty retailer told Modern Retail in October 2025 that the feature would have needed real-time inventory visibility, coupon and promotion handling, customer data collection, loyalty program integration and in-store pick-up support before it could function as a credible sales channel. None of those integrations materialized.
The commercial logic behind retailer resistance runs deeper than technical gaps. As Marshal Cohen, Chief Retail Adviser at market research firm Circana, framed it: ceding the checkout to an AI intermediary means ceding cross-sell opportunity, loyalty accumulation and customer data – and reducing the brand to a commodity supplier processing replenishment orders.
“Why would I want to give someone else control of my customer base,” Cohen said, “and then I become a commodity-based business?”
OpenAI reframes the model: ACP-powered discovery, merchant-owned checkout
OpenAI’s March 24 statement acknowledged the limitations directly. The company wrote that the initial version of Instant Checkout “did not offer the level of flexibility that we aspire to provide,” and announced a shift toward ChatGPT as a discovery environment where merchants retain full control of the transaction.
The new architecture centers on the ACP (Agentic Commerce Protocol) as an open infrastructure layer enabling richer, more visual product browsing within ChatGPT – product comparison, image-based search, side-by-side pricing – while routing the purchase itself to each merchant’s own platform. Merchants are now invited to build dedicated apps within ChatGPT rather than operating within a shared checkout template.
Most retail executives are tempering expectations about the pace of behavioural change. In the near term, the operational priority for most retailers is to ensure that product detail pages and site architecture are readable and indexable by AI platforms – the equivalent of search engine optimization for the generative AI discovery layer.
What this means for sporting goods brands and retailers
The first wave of AI commerce experimentation has produced a clear conclusion for the sporting goods industry.
AI agents work best as a discovery layer, not a transaction layer. For brands and retailers evaluating AI platform integrations, the practical implication is that product search, comparison and pre-purchase research are the areas of genuine near-term value. These tools can drive referral traffic, surface products to high-intent shoppers and shorten the research phase.
Checkout, loyalty relationships and customer data must stay within the brand’s own ecosystem. Sporting goods retailers whose commercial models depend on retaining repeat customers through loyalty programmes, seasonal promotions and omnnichannel engagement face the same risks Cohen identified if those touchpoints are handed over to a third-party AI layer: commoditisation and reduced pricing power.
Agentic commerce adoption is gradual enough that brands have time to build their own integration strategies. The window for establishing strong product visibility within AI search environments, however, is opening now.
