Can an AI Agent Buy From Your Website?

Another week in digital, another headline about artificial intelligence turning the retail world upside down. You’ve likely seen the buzz across LinkedIn, read the tech columns, and wondered if it’s just another passing novelty. OpenAI has rolled out Instant Checkout inside ChatGPT, backed by payment giant Stripe and powered by something called the Agentic Commerce Protocol (ACP).

Suddenly, your customers aren’t just asking an AI for gift ideas. They are letting the AI complete the purchase for them directly within the conversation.

Before you peer nervously at your website analytics or reach for the emergency biscuit tin, let’s have a proper chat about what ‘agentic commerce’ actually is, how it works under the bonnet, and why preparing your website today makes your development strategy look two years ahead of the curve.

What’s the big deal with ChatGPT Instant Checkout anyway?

For nearly thirty years, e-commerce has been engineered around one core assumption: a human being sits in front of a screen, browses a catalogue, clicks ‘add to cart’, and manually types in their payment details. Everything from responsive web layouts to promotional banners and abandoned cart pop-ups exists to guide human eyes down a visual sales funnel.

Instant Checkout flips that entire model on its head.

Launched by OpenAI and Stripe, Instant Checkout allows ChatGPT’s 700 million weekly active users to discover products, evaluate options, and buy them right inside the chat window. There is no redirect to an external web page, no loading screen, and no multi-step checkout form. The feature has already launched with Etsy sellers and is rolling out across more than a million Shopify merchants, including major global brands like Glossier, SKIMS, Spanx, and Vuori.

Instead of browsing your website, a customer says: “Find me a waterproof cycling backpack under £100 with a padded laptop compartment,” and ChatGPT surfaces the best match. If the merchant supports Instant Checkout, a neat ‘Buy’ button appears. The user taps it, confirms their delivery address and saved payment method, and the sale is completed on the spot.

The AI agent has officially become the shopper.

How does an AI agent actually buy something? (The clever bit)

You might be picturing an AI bot opening an invisible web browser, frantically clicking your navigation menu, and filling out checkout inputs like a robotic ghost.

While some early tools used that kind of fragile ‘screen-scraping’ automation, it’s notoriously clunky. Change a button class, launch an A/B test, or tweak your CSS, and traditional browser bots fall over. Plus, running complex browser automation costs a small fortune in computing power and frequently gets stopped in its tracks by security firewalls.

Instant Checkout works through a much smarter, protocol-driven architecture called the Agentic Commerce Protocol (ACP). Co-developed by Stripe and OpenAI as an open-source standard under the Apache 2.0 license, ACP gives AI agents and online shops a common language to talk to each other directly via Application Programming Interfaces (APIs).

Here is how the transaction actually unfolds behind the scenes:

Direct API Communication: Rather than reading your visual HTML pages, ChatGPT speaks straight to structured backend endpoints to check product availability and create a checkout session.

Authoritative Calculations: The AI agent never guesses shipping rates, discounts, or taxes. Your backend recalculates the authoritative totals and returns the verified price to the agent.

Shared Payment Tokens (SPTs): To keep financial details safe, Stripe introduced the Shared Payment Token. The AI model never sees or stores sensitive card numbers. Instead, Stripe issues a single-use, cryptographically scoped token tied strictly to your merchant ID and capped at the exact basket amount.

You Remain the Merchant of Record: ChatGPT does not sell the product; you do. The order flows into your existing backend, where you approve the transaction, capture payment via your payment provider, and handle dispatch and customer care exactly as you always have.

Why your current website might slam the door in their face

This sounds brilliant in theory, but here is the catch: most traditional e-commerce websites today are actively built to stop AI agents from doing business with them.

For years, digital security teams have operated on a simple rule: human traffic is good, bot traffic is bad. If a non-human visitor arrives at your checkout, your Web Application Firewall (WAF) or bot management service usually serves up an interactive CAPTCHA, a JavaScript challenge, or a flat-out block.

However, automated traffic now represents well over half of all web requests. If your security perimeter treats every AI agent like a malicious scraper, you will end up blocking genuine buyers with ready wallets. We have written before about making your site work for humans and AI crawlers at once, and the same balancing act applies here.

To solve this, infrastructure providers and payment networks like Visa and Mastercard have introduced frameworks like Web Bot Auth (built on the RFC 9421 HTTP Message Signatures standard). Verified AI agents can cryptographically sign their requests to prove they represent a legitimate buyer, allowing your firewall to wave them through while continuing to block bad actors.

Furthermore, it’s worth taking a quick peek at your website terms and conditions. Many standard UK e-commerce terms carry boilerplate clauses prohibiting any automated access or robotic ordering tools. Updating those technical policies ensures legitimate agent purchases aren’t left in legal limbo.

What does getting ‘agent-ready’ actually look like?

Preparing for agentic commerce doesn’t mean scrapping your brand or abandoning human customers. It is about ensuring your e-commerce infrastructure can cater to human eyes and programmatic systems with equal finesse.

When our development team designs future-proof e-commerce builds at Platform81, we focus on four core pillars:

1. Structured, Machine-Readable Product Data

AI agents do not care about flowery marketing copy or pretty lifestyle photography. They evaluate concrete attributes: materials, dimensions, precise specifications, real-time stock levels, and compatibility. Upgrading your Product Information Management (PIM) and structuring catalogue feeds via clean schemas ensures an AI won’t skip over your inventory when searching for exact customer matches. The discipline is much the same as keeping a Google Shopping feed clean and complete. It is also the same groundwork that underpins strong e-commerce SEO, so the effort pays twice.

2. Headless and Composable Architecture

If your content, styling, and checkout logic are tangled up into one rigid template, integrating with new conversational channels becomes an expensive headache. With a headless build, your backend business logic, covering pricing, stock, cart creation, and payments, is decoupled from your front-end design. Connecting an agentic gateway or supporting protocols like ACP becomes a clean, lightweight extension rather than a complete rebuild. It is the approach we now default to for any serious e-commerce website development project.

3. API-First Checkout Endpoints

To let external agents interact with your store smoothly, your checkout pipeline needs to support stateless, RESTful session management. Exposing endpoints that can create a session, calculate dynamic delivery options, and accept secure delegated payment tokens gives you total control over how external platforms connect to your store. Much of this is already within reach on mainstream platforms, and a well-planned approach to Shopify website development or WooCommerce development exposes these endpoints without a ground-up rebuild.

4. Smart Edge Governance

Your website hosting and CDN configurations must be tuned to separate malicious scrapers from authenticated AI buyers. By implementing modern signature checks and adaptive rate-limiting, you ensure safe programmatic purchasing without compromising security.

The Platform81 Verdict: Thinking Two Years Ahead

Is every shopper going to abandon standard online storefronts tomorrow? Of course not. Visual discovery, strong brand storytelling, and thoughtful UX and UI design will always have a vital place in digital retail.

However, conversational discovery is rapidly transforming into a powerful commercial sales channel, which is exactly why generative engine optimisation has climbed up the agenda. The brands that succeed over the next two to three years will be those whose digital platforms can effortlessly service both an individual browsing on their mobile and an AI agent completing a task on their behalf.

The best part? The technical investments needed to make your website agent-ready, namely clean code, headless flexibility, structured data, and rapid APIs, are the exact same foundations that good conversion rate optimisation relies on for real people today. It isn’t speculative guesswork; it is good, robust digital engineering.

If you want to ensure your e-commerce platform is built to handle the future of online shopping, our web development and digital marketing teams are ready to help. Get in touch with Platform81 today, and let’s talk about building an e-commerce platform that stays well ahead of the competition.