People Can Now Buy Inside ChatGPT: What the Agentic Commerce Protocol Means for Indian D2C

A sale that closes without the shopper leaving the chat

Picture the full journey. Someone opens ChatGPT, types "best magnesium supplement for sleep under ₹1,500," reads the back-and-forth, and lands on a product they like. In the older model, they'd click out to your website and check out there. In the newer one, there's a buy button right inside the conversation. They tap it, pay, and they're done. They never saw your homepage, your bundle upsell, or your cart.

That's Instant Checkout, and it runs on the Agentic Commerce Protocol, an open standard OpenAI built with Stripe. It started in the US in 2026, building on the single-item Instant Checkout OpenAI first introduced in late 2025. It isn't in India yet. But the direction it points, discovery and payment collapsing into one AI surface, is worth understanding now rather than the week it arrives.

We've written before about discovering in AI but buying on your own site. This is the other branch: buying in the AI itself. The mechanics, the costs, and the control tradeoffs are different enough to deserve their own look.

How "Buy it in ChatGPT" actually works

Under the hood it's less magical than it sounds, which is good news, because it means you can prepare for it concretely.

The checkout steps

The Agentic Commerce Protocol defines a small set of checkout steps, create, update, complete, and cancel, that an AI assistant can call against a merchant. Payment is handled through Stripe's Shared Payment Tokens, which are single-use and time-limited, so neither ChatGPT nor the merchant handles raw card details. The protocol is open and Apache 2.0 licensed, so it isn't locked to a single platform; it can be implemented as a REST API or an MCP server. That openness is the part worth noting, because an open standard tends to get adopted by more than one assistant, which is how a "ChatGPT feature" becomes an industry pattern.

The feed is the entry ticket

To appear, a brand submits a product feed, compressed files refreshed as often as every 15 minutes, carrying the fields you'd expect: a product identifier like GTIN, UPC, or MPN, the title, price, availability, seller name, and links to your return, privacy, and terms pages. Checkout also requires that your products are eligible to surface in search in the first place. No feed, no presence. If that list looks familiar, it should: it's close to what Google Merchant Center has always wanted, which means the work isn't wasted even if you never switch on in-chat checkout.

The money part

On cost, OpenAI charges Shopify merchants a reported 4% fee on completed transactions, with a free-trial window on early orders. That fee sits on top of your normal payment processing, so the combined take on a ChatGPT sale is high-single-digit, not the 2 to 3% you're used to on-site. OpenAI has said the fee doesn't influence rankings, so paying more doesn't buy you placement. You're paying for reach and convenience, not for a better slot.

What's live, what's not, and the India reality check

It's easy to read a US launch headline and assume it's everywhere. It isn't.

For an Indian D2C founder, the honest summary is: you can't switch this on today, and you shouldn't restructure your business around a feature that hasn't reached your market. What you can do is make sure that when the door opens, you're not starting from zero. The groundwork is identical to work that pays off right now.

The one case where it's already relevant

There's a narrow exception worth naming. A handful of Indian D2C brands already run a US arm, a US-registered entity, a US Shopify storefront, and US fulfilment, to sell to American customers. If that's you, the "US sellers, US buyers" boundary isn't theoretical: your US storefront could become reachable inside ChatGPT for US shoppers while your India business waits. For everyone selling only into India, it stays a roadmap item. Knowing which bucket you're in tells you whether this is a this-quarter question or a watch-and-prepare one, and plenty of founders assume they're in the second bucket without checking whether their US setup qualifies.

Discover-in-AI versus buy-in-AI: the real decision

When this does reach India, the choice won't be technical, it'll be strategic. Letting a customer buy inside ChatGPT means a frictionless path for the shopper and reach into a surface where people increasingly start their search. It also means you give up the checkout: no cart, no upsell, no order-bump, less first-party data, and a fee that eats margin.

There's no single correct side. A ₹499 impulse SKU with no meaningful upsell might be perfect for in-chat checkout. A ₹2,000 skincare routine that lives or dies on bundling and subscription almost certainly belongs on your own store. Most brands will end up doing both, by product, and the ones who've thought it through will choose deliberately instead of defaulting.

What to do now, even though it's not in India

The useful part: nearly all the preparation overlaps with being found in AI Overviews and Google's AI Mode, which already affect Indian shoppers. So this isn't speculative work.

Fix the feed first

Start with your feed. A clean, complete product feed with real GTINs, accurate prices, live availability, and valid return, privacy, and terms URLs is the entry ticket to agentic commerce and a quiet ranking factor across AI shopping surfaces generally. Many Indian D2C feeds are patchy here, missing identifiers, carrying stale stock status, or pointing to a returns page that 404s. None of that blocks a normal website sale, so it goes unnoticed, right up until a machine is the one reading it.

Make your brand quotable

Then make your brand quotable to machines. Clear product descriptions, specs stated plainly, honest comparisons, and FAQ content that answers the questions a shopper actually asks. AI assistants recommend what they can confidently summarise, and they struggle with vague, adjective-heavy copy that never states a size, a material, or a use case. If building that discovery-and-feed readiness is where you want help, that's the kind of foundation our growth marketing work is built to set up, so you're visible when buying moves into these surfaces.

Run the margin math

Finally, get your margin math honest. If a 4% platform fee plus processing lands you near 7% all-in on a sale, you need to know which SKUs can absorb that and still contribute. Brands running thin contribution margins on COD-heavy orders will feel it most, and that's a decision to make with a spreadsheet, not a gut call, before the feature ever reaches checkout here. If you already track per-SKU contribution margin, this is an afternoon's work. If you don't, now's the time to build that view, because deciding it under launch-week pressure never goes well.

The quiet shift worth watching

The headline is a buy button in a chatbot. The real story is that the checkout, long the one piece of the journey a D2C brand fully owned, is becoming something you might rent from an AI surface. That's not automatically bad; reach and convenience are real. But it's a tradeoff, and tradeoffs reward the brands that see them coming.

India isn't on the list yet. Use that lead time. Fix the feed, sharpen the content, run the margin numbers, and keep a clear view of which products you'd ever want sold outside your own store. When agentic checkout does arrive, that preparation turns a scramble into a simple switch.