AI Agents That Shop for Customers: Is Your Website Ready?
A customer tells ChatGPT they need a birthday gift under $80, and ChatGPT does not just suggest where to look. It finds a specific product, checks the price, and completes the purchase, all inside the chat, without the customer ever opening your website. That is agentic commerce, and it has moved from a conference keynote to a live feature in the space of a year.
For most Australian businesses it is not yet something to integrate. It is something to understand now, so the groundwork is already in place when it reaches here properly, because the groundwork turns out to be exactly what already helps you win in AI search today.
What “agentic commerce” actually means
Agentic commerce is shopping where an AI agent does the discovery, comparison and purchase on a customer’s behalf, instead of a human clicking through your site. The customer still approves the purchase and pays, but the AI, not the person, is the one browsing your catalogue, checking your price and initiating the transaction.
Two protocols are driving this right now:
- OpenAI’s Agentic Commerce Protocol (ACP), built with Stripe, powers Instant Checkout in ChatGPT, launched in October 2025 with Etsy and Shopify as early partners. The merchant stays the merchant of record; Stripe’s Shared Payment Token lets ChatGPT pass a payment through without ever seeing the customer’s card details.
- Google’s Universal Commerce Protocol (UCP), announced in January 2026, lets shoppers complete a purchase directly inside AI Mode in Google Search or the Gemini app, paying with Google Pay while the retailer remains the merchant of record. Google has said onboarding is rolling out gradually across the US, Canada and Australia, which is the clearest signal yet that this is heading our way, even if most local businesses are not eligible today.
Both protocols solve the same problem from different ends: they give an AI agent a standard, secure way to see your product data and hand off a payment, without you having to build a bespoke integration for every AI platform.
Why this matters even though most businesses cannot plug in yet
Early access to both protocols is limited to large retailers and major platforms. Shopify made its storefronts discoverable to ChatGPT, Microsoft Copilot, Google AI Mode and Gemini by default for eligible merchants in March 2026, and Google’s agentic checkout list so far leans on household names like Nike, Target, Wayfair and Shopify brands such as Fenty. If you run a local Australian store or service business, you are not being onboarded this quarter.
It is also worth being honest about how uneven the rollout has been. OpenAI’s first version of Instant Checkout had barely a dozen Shopify merchants actually ship against it before the feature was pulled back and reworked in March 2026, even as the underlying protocol kept developing. The lesson for a small business is not “this is hype, ignore it.” It is “the checkout mechanics are still shifting, so do not bet your roadmap on today’s specific integration.” Bet instead on the layer underneath it, which is far more stable: can an AI agent actually understand what you sell, at what price, with what availability?
That layer is the same one that already decides whether ChatGPT, Gemini, Perplexity and Google’s AI Overviews mention your business at all when someone asks a shopping question today, agentic checkout or not.
What AI shopping agents need from your site right now
Whether or not a customer ever completes a purchase inside an AI chat, agents are already reading your site to answer product and availability questions. Three things determine whether they can do that accurately:
1. Structured product data, not just product pages
An AI agent cannot reliably infer your price, stock status or delivery promise from marketing copy. It needs that information in a format it can parse without guessing. Product and Offer schema (JSON-LD, the same structured data behind rich results) spells out the price, currency, availability and condition explicitly. If you have already implemented LocalBusiness and Organization schema for AI search and GEO, extending it to Product and Offer on every item you sell is the natural next step, not a separate project.
2. Accurate, current information everywhere it appears
Agentic commerce runs on trust between the AI platform and the merchant. An AI will not route a transaction, or recommend a business in a normal answer, to a site whose stock levels, prices or shipping timelines are stale or inconsistent with what shows at checkout. The same consistency discipline that matters for your Google Business Profile and local listings now extends to every product fact a bot might read.
3. Clear, machine-readable policies
Returns, shipping timeframes and delivery areas need to be stated plainly, not buried in a PDF or a vague “contact us for details.” An AI agent assembling an answer, or in time, a transaction, needs a definitive answer it can quote. Ambiguity that a human customer might shrug off and ring you about is a dead end for an agent, which will simply default to a competitor with a clearer answer.
None of this requires a checkout protocol integration. It is the same foundational work covered in our guide to structuring content for AI search: make your business legible to machines, consistently, everywhere.
What to actually do this quarter
- Audit your product schema. If you sell online, check whether
ProductandOffermarkup exists and matches your live prices and stock. Most Shopify, WooCommerce and BigCommerce themes support this natively; the gap is usually that it was never switched on or has drifted out of date. - Tidy your product feed. The same feed that powers Google Shopping is increasingly the feed AI platforms will read. Incomplete titles, missing GTINs or wrong availability flags hurt both today.
- Write policies for a machine to read, not just a human to skim. A dedicated returns and shipping page with explicit timeframes, not a line buried in your FAQ, is the difference between an AI confidently recommending you and skipping you.
- Watch the Australian rollout, do not chase it. When Google or OpenAI opens agentic checkout to Australian merchants more broadly, having clean structured data already in place means you are a fast “yes” instead of a six-month integration project.
A furniture retailer we worked with learned a related lesson the hard way: strong traffic means nothing if the buying experience has friction baked in. Fixing structure and checkout together, not one before the other, is what moved the needle in our ecommerce SEO and checkout case study. The same principle applies here. An AI agent is simply the newest, least patient customer at your checkout.
The bottom line
Agentic commerce is real, moving fast, and still mostly closed to small businesses, so there is no checkout protocol to rush and integrate this week. But the readiness work, accurate structured data, a clean product feed and clear policies, is the same work that already earns you visibility in AI search today. Do it now and you are ready whenever ChatGPT or Gemini starts sending Australian customers your way; skip it and you are invisible to both the AI agents answering questions today and whatever shops for your customers tomorrow.
If you want a clear read on how AI tools currently see your business, and what it would take to be genuinely ready for agentic commerce, explore our AI Search and GEO services or book a free strategy call.
