The Death of the Buy Button?

What OpenAI’s Checkout Retreat Really Means for Retailers

Six months. That’s all it took for one of the most hyped experiments in AI commerce to quietly fold.

When OpenAI launched Instant Checkout inside ChatGPT in September 2025, the promise was bold. Turn the world’s most-used AI assistant into a seamless shopping destination. No tabs. No redirects. Just conversation to conversion in a single flow.

The partner ecosystem looked serious too. Shopify, Etsy, Walmart, Stripe.

On paper, it was the future.

By March 2026, it was gone.

Not with a bang. With a quiet product note and a pivot.

What Actually Happened

The headline numbers tell you most of what you need to know.

Out of millions of Shopify merchants, fewer than a dozen ever went live with the ChatGPT buy button. That’s not a scaling issue. That’s a signal. Users were happy to browse inside ChatGPT. They asked questions, compared products, uploaded images and explored options.

But when it came time to buy, they left.

They went back to what they trust. Their Amazon accounts. Their saved Shopify stores. Their loyalty ecosystems.

And underneath that behaviour sat some very real operational gaps:

• No robust system for collecting and remitting sales tax

• No multi-item cart support

• No meaningful integration with promo codes or loyalty programmes

• No reliable, real-time inventory sync across large catalogues

In other words, all the hard parts of commerce were still.. hard.

Analysts at TD Cowen called the retreat a "stunning admission." OpenAI’s own explanation was more polite but revealing. The product “did not provide the level of flexibility” they were aiming for. Translation: building a checkout platform is a full-time job. Doing it as a side quest to an AI product is a different level of complexity entirely.

The Bit Everyone Should Be Paying Attention To

This is where most takes get it wrong.

This is not the death of AI commerce. And it’s definitely not proof that consumers don’t want AI in the buying journey. What it is is clarity. AI isn’t the checkout, it’s the layer before it.

Consumers are already using ChatGPT for discovery at scale:

• “Find me something like this” (image uploads)

• “What’s the best option under £200?”

• “Compare these three products”

That behaviour is accelerating. And at the exact same time OpenAI killed checkout, they doubled down on discovery. Better search. Richer comparisons. Visual inputs. Broader access. The shift is obvious:

From AI as the checkout → to AI as the path to checkout.

What This Means for Retailers (Right Now)

This is where it gets practical.

1. Your product data just became a frontline asset

If AI is deciding what to show, your data decides if you exist. Titles, descriptions, structured attributes, imagery. All of it.

This is not traditional SEO, this is something closer to Answer Engine Optimisation.

If your product data is weak, you’re invisible before the customer even clicks.

2. Your checkout is your moat

There’s a reason users didn’t convert inside ChatGPT. Trust.

Saved cards. Loyalty points. Order history. Delivery preferences. These aren’t features, they’re comfort. And they’re incredibly hard for a third party to replicate. OpenAI’s retreat reinforces something most retailers already felt: The last mile belongs to you. So make it count.

3. The real battleground is the app layer

Instead of owning checkout, OpenAI is shifting toward integrations.

Think connected experiences with:

• Instacart

• Walmart (via its own AI assistant)

• Travel platforms like Expedia

Walmart, in particular, has already pivoted. Moving away from embedded checkout to a deeper, app-driven experience with account linking, loyalty integration, and native payments.

That’s the pattern.

AI drives intent. Your ecosystem captures conversion.

4. Agentic commerce is coming… in a bit

There’s a temptation to write this off as “AI shopping doesn’t work.” That’s too simplistic.

The reality is the infrastructure isn’t ready:

• Tax compliance

• Fraud handling

• Inventory synchronisation

• Loyalty systems

These are solvable problems. But they take time, coordination, and standards. Importantly, retailers who start preparing now will be ahead when it clicks into place.

The Bottom Line

OpenAI didn’t abandon commerce, they abandoned a version of it that arrived too early.

• The discovery layer is already here and growing fast.

• The conversion layer is being rebuilt more carefully.

So the question isn’t “Should we care about AI commerce?”, that decision has already been made for you. The real question is Are you easy for AI to find, recommend, and send traffic to?

Because the buy button didn’t disappear.

It moved.

And it now lives firmly back where it probably always belonged.

On your site.

Where Cabiri Sees This Going

This is exactly the kind of shift we’ve been building toward.

Not AI replacing commerce platforms, AI amplifying them.

The retailers who win here won’t be the ones chasing the next shiny integration.

They’ll be the ones who:

• Treat product data as a strategic asset

• Invest in fast, frictionless checkout

• Build flexible architectures that can plug into whatever comes next

Composable, Shopify, hybrid. The label matters less than the outcome.

Can you adapt quickly when the path to purchase changes?

Because it just did.

If you’re thinking about how AI-driven discovery fits into your current stack, or where to invest next, let’s have a conversation.Analysts at TD Cowen called the retreat a "stunning admission."