The Pivot OpenAI Doesn’t Want to Call a Pivot
The original pitch was seductive: a shopper types “I need linen trousers for a Greek wedding” into ChatGPT and buys them, right there, inside the chat window. No merchant site. No checkout redirect. OpenAI ran that playbook for six months with about thirty launch merchants. Then they switched Instant Checkout off.
The replacement model inverts the logic. Discovery still happens in ChatGPT. The conversation, the recommendation, the product cards — all native to the assistant. But the moment you click buy, ChatGPT opens the merchant’s real storefront in an in-app browser. The retailer keeps the cart, the data, the payment processing, the customer relationship. ChatGPT becomes the top of the funnel, not the funnel itself. Shopify flipped a different switch weeks later, and the whole category moved with it.
OpenAI is calling this an “evolution.” Everyone else is calling it what it is.
Why Instant Checkout Failed
Three reasons, and they are instructive for anyone building agentic anything:
Scale was a fantasy. Roughly thirty launch merchants went live in six months. Shopify alone has millions. A checkout experience that only works with 0.01% of the internet’s stores is a demo, not a product.
The data was wrong. Instant Checkout leaned on web scraping for pricing and inventory. Prices drifted. Stock-outs weren’t caught. Agents confidently sold items that didn’t exist. Nothing destroys trust in an AI assistant faster than a confirmed order for a ghost SKU.
Merchants resented it. Retailers spent a decade building direct-to-consumer relationships to escape Amazon’s customer-data black hole. Handing that relationship back to a new intermediary — one that owned the checkout, the receipt and the re-engagement pixel — was not a trade most brands were willing to make. The launch cohort was small because the economics were hostile.
Agentic Storefronts fix all three at once.
Agentic Storefronts: How It Actually Works
On 24 March 2026, Shopify activated ChatGPT Product Discovery for every store on the platform. The mechanic is clean:
- A shopper asks ChatGPT for something — “comfortable travel joggers under £120.”
- ChatGPT returns structured product cards, pulled from Shopify’s live catalogue feed, with accurate pricing, accurate inventory and accurate shipping windows.
- The shopper taps a product. ChatGPT opens the merchant’s storefront inside an in-app browser.
- Checkout completes on the merchant’s own Shop Pay (or Stripe, or whatever they run). The retailer keeps everything: the transaction, the customer record, the marketing permission.
The merchant didn’t have to integrate anything bespoke. If you sell on Shopify, you’re already in. ChatGPT went from thirty merchants to millions overnight — not by scaling its own checkout but by deprecating it.
The Agentic Commerce Protocol (ACP) Explained
Underneath the consumer experience sits something more important: the Agentic Commerce Protocol. ACP is the open specification that lets AI agents and merchants transact without bespoke per-agent integrations. Think of it as the HTTP of AI shopping — a shared grammar for “here is a product, here is its stock, here is how an agent checks out on the merchant’s own infrastructure.”
OpenAI and Shopify co-authored ACP. Stripe is a participant. The spec is open, which means it isn’t just a ChatGPT mechanic — any AI agent that implements ACP can discover and transact against any merchant that supports it. That’s the entire point. Closed checkout experiments don’t scale. Open protocols do.
ACP vs Google’s UCP: The Protocol War
Google is running a parallel play. Its Universal Commerce Protocol (UCP) is the Google Shopping and AI Mode equivalent — same shape, different sponsor, different gravitational pull. Both specs aim to do the same thing: let any AI agent transact against any merchant. They will almost certainly converge at the edges, because no merchant wants to implement two incompatible integrations for the same job.
The short version for anyone building a commerce brand: assume your catalogue will need to be machine-readable for at least two protocols by end of 2026. The winners will be the merchants whose product data, pricing, inventory and policies are already structured enough to plug into either one without a six-month engineering project.
What Glossier, SKIMS and Spanx Get That Most Merchants Don’t Yet
The launch partner list reads like a shortlist of brands that are unusually sharp about distribution: Glossier, SKIMS, Spanx, Vuori. These are companies that have already been through one platform transition — the Instagram-to-TikTok moment, the paid-social-to-creator moment — and learned a specific lesson: be first on the new surface, because the top of that surface gets priced for free and the bottom gets priced like legacy ad inventory.
They understand three things most merchants haven’t clocked yet:
- Being discoverable in ChatGPT is a distribution channel, not a tech project. It needs merchandising, copy, bundles and landing-page strategy — the same craft they pour into their homepage.
- Product content is the new ad creative. An AI agent chooses which product to surface based on description, reviews, images and structured data. Sloppy PDPs don’t rank in ChatGPT any more than thin content ranks on Google.
- The first million conversations are a land grab. Early movers train their category’s default answer. If “best shapewear” resolves to SKIMS inside ChatGPT today, every later entrant is paying to dislodge a learned association.
What This Means for Marketing Teams
If you took one thing from the original version of this post, it was probably “AI is going to change how people shop.” That was true but vague. The 2026 version is sharper, and it directly rewires how marketing teams operate.
Discovery for millions of products is now happening inside an AI agent. That agent reads your content. It reads your product data. It reads your reviews, your policies, your structured metadata, your blog posts, your press coverage. It does not read your billboard, your podcast ad or your print spread. It does not care about your paid social retargeting unless it can parse the underlying asset.
Which means the job of a marketing team has quietly mutated. You are not just writing for humans, or even for Google — you are writing for the agent that is about to become the primary discovery surface for a generation of shoppers. Your content has to be AI-readable. Your product data has to be clean enough for a protocol. Your brand voice has to hold up when an AI agent summarises you in one sentence to a stranger.
Most marketing teams are not set up for this. They’re still running a 2019 stack: a content agency, a paid media partner, an SEO consultant, a freelancer for everything in between, and fourteen spreadsheets holding it together. That stack cannot move at the speed an agentic surface demands. That is why Anjin exists: a Marketing Operating System built for the agentic-commerce era.
Anjin: The Marketing Operating System for the Agentic Commerce Era
Anjin is the Marketing Operating System — one platform that runs your content generation, channel distribution, SEO, product data optimisation, brand consistency and campaign planning inside a single agent-native workflow. It’s built for exactly the shift Agentic Storefronts represent: a world where your marketing has to be machine-readable, continuously updated and ready to ship the day a news beat breaks.
What Anjin replaces:
- The content agency writing your PDPs, blog posts and brand pages
- The SEO consultant chasing rankings that now split across Google, ChatGPT and AI Mode
- The paid media planner who used to brief, test and report across channels
- The coordination layer — the Slack threads, Notion docs, spreadsheets — holding it all together
- The £8–15k/month you’re paying to keep that stack alive
What Anjin does that none of them can:
- Writes content that is structured for both human readers and AI agents by default
- Ships the day a protocol moves, not the quarter after
- Learns your brand voice in hours, not months
- Runs 24/7 against the surfaces where discovery is actually happening now
If ChatGPT is the new top of funnel for millions of merchants, your marketing team needs to run at ChatGPT’s clock speed. That is what we built Anjin to do — and the £888 lifetime license means you can lock it in before the early-access cohort closes.
The £888 Lifetime License — Offer Closing Soon
Lifetime access to Anjin for a one-time payment of £888. Not a subscription. Not a seat. Not a trial. One payment, unlimited use, for as long as Anjin exists.
The average marketing team spends £888 in about three working days on tooling, freelancers and coordination software. You’re buying the platform that replaces most of it — once.
This price will not be offered again once we close our early-access cohort.
Claim your £888 Anjin lifetime license →Founders, agency owners and in-house marketers — this is how you run marketing at AI speed without the team, the burn, or another year of waiting.
Sources: Modern Retail, CNBC, Shopify Newsroom, Shopify Newsroom, Shopify Newsroom, OpenAI, Ask Phill, The Keyword
