What brands need to know about ChatGPT’s new ‘buy’ button
With the recent announcement of ChatGPT’s integrated “buy” function (in the US only for now), Leoprd’s Celia Harding looks at how brands need to prepare.

Leoprd founder Celia Harding
Stripe and OpenAI have just announced that ChatGPT users can now buy products directly through AI conversations, powered by Instant Checkout, a new Agentic Commerce Protocol (ACP), and shared payment tokens.
Being able to prompt and pay means we’ve firmly entered the era of the answer economy, where ChatGPT decides which brands get bought or buried. With limited real estate in AI responses, being algorithmically invisible now means being commercially invisible.
Recent research based on a sample of 700m ChatGPT users provides evidence of this shift: 52% of AI interactions involve ‘asking’ — users want decision support, not just information, as AI platforms become the world’s most trusted advisors.
Enjoying Mumbrella? Sign up for our free daily newsletter.
Historically, consumers filtered through pages of Google’s paid and unpaid results to find information and make educated purchases. Now ChatGPT is doing the curation and culling on consumers’ behalf, providing recommendations that more people are trusting. Positive AI interactions now outnumber negative ones 4:1, up from 3:1 last year.
The story is what comes next, and most brands aren’t ready.
Chat to checkout: How can a brand list its products?
You need to apply to OpenAI to connect your merchant feed (sign up here if you haven’t already)
Merchants are ranked based on availability, price, quality, whether they are the maker or primary seller of that item, and whether Instant Checkout is enabled.
This information needs to be structured to be easily parsed by agents. The product feed specification defines how merchants share product data with OpenAI so ChatGPT can accurately surface their products in search and shopping experiences.

ChatGPT’s Instant Checkout feature
Beyond direct product feeds, brands need to optimise for AI visibility more broadly to appear in recommendations. Our LEO Report: Building Trust, Authority and Visibility in the Age of AI found that over 62% of brand citations on LLMs came from earned or third-party signals — specifically editorial coverage, reviews and awards. Not paid ads. Owned content accounted for 20% of citations.
Agentic commerce reinforces this dynamic. ChatGPT doesn’t rank you based on what you say about your products; it evaluates external signals of quality and credibility. You can’t buy placement — you have to earn your place in the limited real estate ChatGPT provides.
With ChatGPT processing 2.5 billion prompts a day and with 85% market share in Australia, these developments require an overhaul of the traditional marketing funnel.
Instant Checkout does not allow merchants to sign customers up for marketing emails from orders placed through ChatGPT. So brands need to find new ways to reach human audiences while optimising for agentic audiences too.
Brands need to start building their integrations immediately. This isn’t a wait-and-see moment — it’s a first-mover advantage scenario. The infrastructure being built today determines who gets recommended tomorrow.
Futureproofing your brand to be part of the conversation
This isn’t sci-fi. This technology has already launched in the US. Australia and other markets won’t be far behind. Those who optimise now will be the ones with a head start in the answer economy.
Design content for decisions, not just discovery. Provide guidance that helps consumers and agents choose, not just browse. Position your brand as the expert advisor, not just the vendor.
Instead of filtering on ecommerce sites by brand, size, or colour, you’ll be able to search by ingredients, sustainability, and ethical considerations. Think about your supply chain, ingredients and formulations now – consumers will be more educated when making purchasing decisions based on what’s important to them.
Right now, users explicitly confirm each purchase step to build trust and adoption of agentic commerce. But we’re witnessing the early stages of true agentic behaviour. In the near future, AI will be making purchasing decisions on behalf of consumers. The question is how fast merchants can adapt to being evaluated by machines, not humans.
Websites must migrate from human-first design to agent-first information architecture. Is your product data structured for AI comprehension? Are you showing up for the natural language prompts shoppers are actually using?
Celia Harding is an AI visibility expert and the founder of language engine optimisation advisory Leoprd (“LEOPRD”).