The window: what US agent checkout means for merchants in Europe and Asia

In September 2025, OpenAI switched on Instant Checkout in ChatGPT for US shoppers. For the first time at consumer scale, an AI assistant could complete a purchase — not link to a store, not fill a cart for review, but buy the thing.

If you sell from Berlin, Warsaw, Singapore or Seoul, that launch did not include you. No US-style agent checkout has reached European or Asian consumers yet. It would be easy to read that as “not my problem yet.”

We’d suggest reading it as a gift.

Asymmetries like this don’t last

The pattern is boringly consistent. E-commerce itself, marketplace platforms, mobile wallets, buy-now-pay-later — nearly every commerce capability launched in one market and reached the rest of the world within a couple of years. The US launch isn’t a fence; it’s a countdown. OpenAI has said it plans to expand internationally; every competitor building AI shopping has global ambitions, because assistants are global products.

What does differ by region is the landing. European payments run through PSD2 and Strong Customer Authentication; checkout flows, consumer rights and data rules genuinely are different here. Whoever brings agent checkout to Europe and Asia will need infrastructure that respects that — which is precisely why it lags the US, and precisely why “wait and copy the American playbook” won’t quite work.

What the window is for

Here’s what makes this moment unusual: you can see the change before it hits your revenue. US merchants got weeks of warning. You have the rare version of a platform shift where preparation is possible.

Getting agent-ready costs a merchant almost nothing today — an app install, a catalogue sync, a test order. What it buys is position: when assistants start completing purchases in your market, they will buy from stores that can accept their orders on day one. Discovery in agent commerce compounds early; the first stores an assistant can reliably transact with become the stores it keeps transacting with.

The alternative is the other queue — the one US merchants without agent checkout are standing in right now, watching a channel grow around them and retrofitting under pressure.

The honest caveat

Nobody can tell you what percentage of your sales will flow through agents in three years, and you should distrust anyone who claims to. What’s verifiable is direction: the technology is live, the protocols are standardising, the operators are expanding. The bet isn’t “this will be X% by 2028.” The bet is “being unable to accept purchases from the fastest-growing buying interface is a bad position” — and that one doesn’t need a forecast.

The window is open. It won’t be the merchants’ fault when it closes — only whether they used it.