OpenAI launched Instant Checkout in September 2025 and sunset it on March 5, 2026 after only about 30 Shopify merchants integrated. It is replacing in-chat checkout with dedicated retailer apps that route buyers to the brand's own site. For brands, the rerouting model is the better deal because you keep the customer, the data, the tax handling, and the post-purchase relationship.
- Instant Checkout let users buy inside ChatGPT; OpenAI sunset it on March 5, 2026.
- It failed on real-time inventory and pricing, unreliable scraped product data causing order failures, and no US sales-tax system.
- The replacement routes shoppers from ChatGPT to the retailer's own website to complete the purchase.
- Rerouting keeps the customer relationship, first-party data, and tax compliance with the brand, not the platform.
In September 2025, OpenAI launched Instant Checkout, letting people buy products without leaving ChatGPT. By March 5, 2026, it sunset the feature. In the months between, roughly 30 Shopify merchants integrated, a small number that tells you the model never reached escape velocity. OpenAI is now building something different: dedicated retailer apps inside ChatGPT that handle discovery in the chat and then route the buyer to the retailer's own website to complete the purchase. ChatGPT is becoming an app store for retailers, not a checkout.
I read this from the merchant's chair because that is the chair these experiments are tested in. Running DTC brands at WIN Brands Group meant living the operational reality that a chat checkout has to solve: inventory that changes by the minute, pricing that moves with promotions, sales tax across dozens of jurisdictions, and returns that have to land somewhere. Instant Checkout under-built every one of those, which is exactly why it stumbled. I wrote the post-mortem in why ChatGPT pulled Instant Checkout. This post is about why the thing replacing it is better for brands, not worse.
The instinct is to read a sunset as a retreat. It is not. The first version asked OpenAI to become a merchant of record it was not equipped to be: holding inventory truth, remitting taxes, owning order failures. The new version asks ChatGPT to do the thing it is actually good at, surfacing the right product to a high-intent shopper, and then hands the transaction back to the brand. That division of labor is healthier, and it lines up with where the value is. Adobe found AI-referred traffic converts 42 percent better than non-AI traffic, so a model that sends that high-intent shopper to your own site is handing you your best-converting traffic on your own terms.
So the headline for a brand is counterintuitive but real: the failed in-chat checkout was the worse deal, and the rerouting model is the better one. You keep the customer, the data, the tax handling, and the post-purchase relationship, the things that actually compound. This post explains why the first version failed, why the second is more durable, and what to build now to win in it. The strategic frame sits inside how agents change discovery and the buy.
What OpenAI changed,
and what the new
model looks like.
The change is a shift in who owns the transaction. Old model: buy inside ChatGPT. New model: discover inside ChatGPT, buy on your own site. Here is the before and after, side by side.
| Dimension | Instant Checkout (sunset) | Retailer apps (now) |
|---|---|---|
Where the buy happens | Inside ChatGPT | On the retailer's own site |
Who owns the customer | Ambiguous, platform-leaning | The brand |
Inventory and price truth | Scraped, often stale | Live, from your store |
Sales tax | No US remittance system | Your existing tax stack |
Merchant uptake | ~30 Shopify merchants | Open to retailer-built apps |
The new model treats ChatGPT the way the web treats a search engine: a place that refers high-intent shoppers to merchants who complete the sale themselves. Merchants can build apps within ChatGPT or simply let the answer engine refer shoppers to their site to order. Either way, the purchase, and everything that comes with it, lands on infrastructure the brand already runs. That is a structurally different relationship than the one Instant Checkout proposed, and a far easier one to operate.
Why Instant Checkout
failed, in three
operational truths.
Instant Checkout did not fail because agentic shopping is a bad idea. It failed because in-chat checkout asked OpenAI to solve three hard operational problems that brands and Shopify already solved years ago, and it solved none of them well.
Notice the pattern. Every failure is something a real merchant operation handles as table stakes. Inventory truth, clean product data, tax remittance, these are not glamorous, but they are the difference between a checkout that works and one that quietly loses orders and creates liability. Onboarding merchants turned out to be arduous, and only about 30 Shopify merchants ever went live, which is the market telling you the juice was not worth the integration squeeze. The deeper lesson, which I keep returning to with founders, is that the operational layer is the moat, and a platform that under-builds it cannot simply will a checkout into existence. That same operational reality shows up in any honest contribution margin teardown, where the unglamorous costs decide whether a sale is actually a good one.
Why rerouting is the
better deal, not the
consolation prize.
It is tempting to see "you have to send them to your own site" as ChatGPT giving brands less. It is the opposite. The rerouting model hands the transaction back to the party that is actually equipped to run it, and keeps the most valuable parts of the relationship with the brand. Walk through what changes.
The purchase happens where the truth lives. Your store knows your real inventory, your real price, your real promotions, and your tax obligations. Completing the buy there eliminates the three failure modes that killed Instant Checkout in one move. No stale data, no missing tax system, no order failures from a scraped view of your catalog. The hardest operational problems disappear because you were already solving them.
"The failed model asked ChatGPT to become a merchant. The new one asks it to be a great front door and lets you stay the merchant. That is the better deal."
And you receive the shopper at their highest intent. Someone who asked ChatGPT for exactly what you sell and clicked through is more qualified than almost any paid click. Adobe's data has AI-referred traffic converting 42 percent better than non-AI traffic, browsing longer, and producing higher revenue per visit. The rerouting model funnels that traffic to your site, where you control the experience, the upsell, the email capture, and the next purchase. If you are set up to receive it well, this is the best-converting top-of-funnel you can get, which is why making your storefront ready for it, the subject of AI shopping assistants on your storefront, is the move.
What you keep that
in-chat checkout
would have taken.
The clearest way to see why rerouting wins is to list what stays with you that an in-chat checkout would have weakened or taken outright. Each of these compounds, which is why platform-owned checkout is a worse long-term trade even when it looks more convenient.
| Asset | In-chat checkout | Rerouting to your site |
|---|---|---|
Customer relationship | Platform-mediated | Direct, yours |
First-party data | Limited, shared | Full, yours |
Email and retention | Hard to capture | Captured at checkout |
Tax compliance | Platform's problem (unsolved) | Your existing stack |
Upsell and bundle | Constrained | Your full merchandising |
Post-purchase and returns | Fragmented | Your normal flow |
The single most important line is the first-party data and customer relationship. In a world where paid acquisition keeps getting more expensive, the brands that win are the ones that own the customer and can sell to them again. An in-chat checkout that intermediated that relationship would have quietly converted you into a wholesaler on someone else's platform. Rerouting keeps you a direct brand. That distinction is worth more than the small friction of one extra click, and it is the same ownership logic I apply to the board conversation on commerce in 2026: the durable value is in the relationship you control, not the channel you rent.
The discovery-in-chat
playbook. Get found,
then receive well.
If the buy moves to your site, the game splits into two jobs: get surfaced inside ChatGPT, and receive the click in a way that converts. Most brands are weak at the first and sloppy at the second. Here is how to be strong at both.
Get found. ChatGPT recommends what it can read and trust. That means complete, structured product data and content that answers the natural-language buying questions your customers actually ask. This is answer-engine optimization, and the mechanics are in answer-engine optimization for commerce. If you are not sure whether you currently appear, the test and fix are in whether your products are winning or invisible in ChatGPT.
Feed clean data. The same data quality that killed Instant Checkout will decide whether ChatGPT recommends you confidently. Accurate price, stock, attributes, and a feed that mirrors live truth. The Shopify-specific build is in how agentic storefronts on Shopify work. Clean data is now a discovery asset, not just an ops nicety.
Receive the click well. When ChatGPT sends a high-intent shopper to your product page, that landing has to be fast, match the recommendation exactly, and make the buy obvious. Capture the email, set up the upsell, and treat the session like the best-converting traffic you have, because per Adobe's numbers, it is. The brands that win the rerouting era are the ones that get discovered and then convert the handoff, not one or the other.
What to build now for
ChatGPT's retailer
app store.
Treat ChatGPT the way you treated Google fifteen years ago and the app stores ten years ago: a major new front door you cannot afford to be invisible on. The work is not exotic, and most of it pays off across every AI surface, not just ChatGPT.
Near term, get the foundation right: structured catalog data, a clean live feed, content that answers buying questions, and a storefront that converts an inbound high-intent click. That foundation is what makes you discoverable and transactable to ChatGPT's rerouting model and to the broader agent ecosystem at the same time. None of it is wasted if a different assistant wins; readability and a strong storefront are universal.
Medium term, watch how the retailer-app model develops and decide whether to build a dedicated app or simply optimize to be referred. For many brands, being the brand ChatGPT names and clicks through to your site will be enough, and it is the lower-lift path. The cooperative, opt-in direction here also lines up with where the law is heading after the first agent-commerce ruling, which I cover in the first agent-commerce legal precedent, and with how the catalog-access fight is playing out in Amazon versus OpenAI. The throughline is consistent: be discoverable, keep the customer, control the buy.
OpenAI's first swing at shopping asked it to be a merchant it was not built to be, and it missed. The second swing asks ChatGPT to be a great front door and lets the brand stay the merchant. That is the better arrangement, and the brands that prepare for it, discoverable in the chat and ready to convert on their own site, get their best-converting traffic on their own terms.
Questions brands ask me
about ChatGPT's shift to
retailer apps.
OpenAI launched Instant Checkout in September 2025 and sunset it on March 5, 2026, after only about 30 Shopify merchants integrated. It failed on three operational fronts: no reliable real-time inventory and pricing, unreliable scraped product data that caused order failures, and no system to collect and remit US state sales taxes. Onboarding merchants was also arduous. The full post-mortem is in why ChatGPT pulled Instant Checkout.
OpenAI is building dedicated retailer apps inside ChatGPT. Instead of buying in the chat, shoppers discover products in ChatGPT and get routed to the retailer's own website to complete the purchase. Merchants can build apps within ChatGPT or simply be referred to their site, like a search engine sends shoppers to merchants. The strategic frame is in agentic commerce for Shopify brands.
Better. The purchase happens on your own site, where your real inventory, pricing, and tax stack already live, which eliminates the three failures that killed Instant Checkout. You keep the customer relationship, first-party data, email capture, upsell, and post-purchase flow. And you receive a high-intent shopper that Adobe found converts 42 percent better than non-AI traffic. Setting your storefront up to receive it well is covered in AI shopping assistants on your storefront.
Make yourself readable and recommendable: complete structured product data, a clean live feed, and content that answers the natural-language buying questions your customers ask. ChatGPT recommends what it can read and trust. The build is in how agentic storefronts on Shopify work, and the answer-engine mechanics are in answer-engine optimization for commerce.
Maybe, but it is not the first move. The foundation, structured data, clean feed, buying-question content, and a storefront that converts a high-intent click, makes you discoverable and transactable across ChatGPT's rerouting model and the broader agent ecosystem at once. For many brands, being the brand ChatGPT names and clicks through to your site is enough. Decide on a dedicated app after the foundation is solid and you can see the traffic.
Want ChatGPT to send buyers to your store, ready to convert?
I was the merchant these checkout experiments get tested on, so I read this pivot through the operational reality of inventory, tax, data, and returns. That is the view I bring to making your brand the one ChatGPT recommends, and the one that converts the click when it arrives.
Start a conversation See the case studies →A note on sources: the September 2025 Instant Checkout launch, the March 5, 2026 sunset, the ~30 Shopify merchants, and the failure reasons (no real-time inventory and pricing, unreliable scraped product data, and no US sales-tax system) are reported by CNBC and Retail Insight Network, with analysis from Forrester. The 42% AI conversion lift is from Adobe Analytics. The read on why rerouting is the better deal is mine, drawn from operating DTC brands.