In September 2025, OpenAI and Shopify announced what seemed like the beginning of a new retail era: Instant Checkout, a feature that would let consumers complete purchases directly inside ChatGPT. No new tab. No separate cart. Just a conversation that ended with a confirmed order. Within months, the companies promised, anyone could ask ChatGPT for Mother's Day gift ideas and buy in a single session.
By March 2026, it was dead. According to reporting by Fast Company's Elizabeth Segran, fewer than 30 of Shopify's millions of merchants ever went live. OpenAI's official explanation cited a lack of "flexibility." The more candid read from the industry: no one was prepared for how technically complex it is to bridge a large language model with the real-world plumbing of commerce — loyalty programs, inventory systems, regional promotions, and the dozens of edge cases that any working checkout must handle.
For CPG brand leaders, the collapse of Instant Checkout is not a cautionary tale about premature hype. It's a signal. The AI companies with the deepest resources on earth tried to shortcut the hard infrastructure work — and couldn't. The work is now happening in earnest, and the choices being made in 2026 will determine which brands are visible, discoverable, and purchasable in the agentic era. The ones that wait for the smoke to clear may find the room already occupied.
Why the First Attempt Failed
The failure of Instant Checkout reveals something important about where agentic commerce actually is — versus where it was being marketed. LLMs are trained on text scraped from the internet, but a product page is fundamentally different from a web page. Real-time inventory, shipping costs, promotional pricing, loyalty points, and regional availability rarely appear in crawlable text. They live in backend systems that LLMs have no native way to read.
OpenAI's commerce lead Neel Ajjarapu acknowledged to Fast Company that the complexity was underestimated: building a checkout "is not enough." You need to account for loyalty points, in-store pickup, basket promotions, "and dozens of features that are specific to the geography, category, and merchant type." The conclusion OpenAI reached — that merchants should own their own checkout — is the right one, but it still leaves the discovery and recommendation layer squarely in AI territory. That layer is where the real competition is being fought.
"Shopify has major egg on their face. OpenAI said, 'We've solved it.' But they literally couldn't even get it live."
— Omar Qari, CEO, Logicbroker, as quoted in Fast CompanyGoogle's Data Moat Is the Story
With OpenAI stepping back from checkout, the question of who becomes the dominant AI shopping interface sharpens into a more fundamental question: who knows the consumer best? This is where Google's structural advantage becomes, in our view, the most significant and underappreciated fact in agentic commerce.
ChatGPT can only personalize recommendations based on what a user shares in a given conversation. Gemini — with its March 2026 rollout of Personal Intelligence — can draw on years of Gmail, Photos, and Drive data from its 750 million monthly active users. As Google's VP and GM of Advertising and Commerce Vidhya Srinivasan explained to Fast Company, the implication is profound: "Once the model has the opportunity to learn about you, it can start at a better point than starting from scratch and expecting that you will tell us everything about you."
"Once the model has the opportunity to learn about you, it can start at a better point than starting from scratch."
— Vidhya Srinivasan, VP & GM Advertising and Commerce, Google, as quoted in Fast CompanyGoogle's Shopping Graph — a real-time product database built over more than two decades — gives it a second moat on top of the first. It knows the user and it knows the inventory. Gap and Ulta Beauty have already enabled full in-chat checkout via Gemini. That list will grow quickly. For CPG brands, the practical implication is this: Google is not just one of several platforms to monitor. It may be the platform that determines whether your product is ever surfaced to a personalized AI shopper at all.
The Protocol War You've Probably Never Heard Of
While the consumer-facing battle plays out in headlines, a quieter war is underway at the infrastructure level — and it may matter more. AI companies have realized they need a new technical standard to make retailers' real-time product data readable by LLMs. Two competing protocols have emerged.
OpenAI and Stripe co-developed the Agentic Commerce Protocol (ACP). In response, Google, Shopify, and a coalition that includes Target, Walmart, Visa, and Mastercard launched the Universal Commerce Protocol (UCP) in January 2026. These are the rails on which agentic shopping will run. The protocol that wins becomes the equivalent of HTTP for AI commerce — invisible to consumers, foundational to everything.
For CPG brands, this isn't an abstract technical debate. Product content that isn't structured to the winning standard will be functionally invisible to agents, regardless of brand equity, marketing spend, or shelf presence. A brand that has spent 20 years building consumer recognition can be sidelined entirely if its product data isn't machine-readable in the way agents require. This is the hidden threat that most CPG teams haven't yet prioritized.
What CPG Brands Should Do Before the Dust Settles
The conventional wisdom in moments of platform transition is to wait for a winner and then optimize. That was a viable strategy in the early days of SEO and social media. It is not viable here. The brands building relationships with AI platform teams, structuring their product data for agent readability, and populating both ACP and UCP-compatible feeds are building an advantage that will be very hard to replicate in 18 months when urgency becomes consensus.
The specific actions that matter now: audit your product content for structured data completeness — not for traditional SEO, but for what agents actually parse (attributes, use cases, pairings, real-time availability). Pursue direct data partnerships with both Google Merchant Center and OpenAI's merchant program. And invest in understanding how your category behaves inside AI-generated recommendations — because the brand that gets cited first in an agent's answer is the brand that captures the purchase.
The race is not decided. OpenAI is not finished — their commerce lead put it plainly: "Every couple months, we just see such massive changes to what our models are able to do. It is impossible for me to predict what's going to happen on what timeline." But the infrastructure is being built right now. The brands that treat this as a spectator sport will find themselves on the wrong side of a transition that looks, in retrospect, as obvious as the shift to e-commerce in 1999.
This analysis draws on reporting by Fast Company's Elizabeth Segran (May 7, 2026), McKinsey Global Institute data, and Pew Research Center. Executive quotes are sourced from Segran's reporting and cited accordingly. Analysis and editorial perspective are original to The Invisible Shelf / TheoryNXT.