Something is shifting in the plumbing of online retail, and most New Zealand business owners have not noticed because it does not show up on their own website. The change is happening one layer down, in the machine-readable data that sits behind every product page. For twenty years the job of an online store was to persuade a human being who had landed on it. That job is quietly being handed to software that never looks at the page at all.
The software in question is the AI shopping agent. When someone asks ChatGPT, Google’s AI Mode or Microsoft Copilot to find them the best value merino base layer under a hundred dollars, the assistant does not open a dozen browser tabs and read the marketing copy. It queries structured product data, compares price and availability across whatever stores have made that data available, and hands back a shortlist. Increasingly it can take the next step and begin the purchase. The customer never sees the homepage a business spent good money designing.
This is not a forecast. Shopify has reported that AI-driven traffic to its stores grew eight times over the year to the first quarter of 2026, and that orders originating from AI searches rose almost thirteen times in the same period, according to its own account of how agentic commerce works. Analysts watching the shift have also noted that shoppers arriving from AI tools are now converting better than ordinary web traffic, a reversal of the pattern from a year earlier when those same visitors were markedly less likely to buy. The people using these tools have moved from curiosity to spending.
Underneath all of this sits a small pile of new standards with unglamorous names. The Universal Commerce Protocol, co-developed by Shopify and Google and put forward at the start of 2026, defines how an AI agent creates a cart, checks out and pays across different platforms through a single integration. A parallel effort, the Agentic Commerce Protocol from OpenAI and Stripe, does much the same for ChatGPT. The detail that matters for a store owner is the direction of travel. A merchant now publishes a machine-readable manifest at a standard location on their domain, product data flows to the assistants as structured feeds, and the shopper completes the sale on the merchant’s own checkout with the store still the merchant of record. The assistant is a new shopfront window, not a new landlord.
What does this mean for a business in Hamilton or Hawke’s Bay selling online? The honest answer depends on how the store was built. Merchants on Shopify are in a comfortable position, because Shopify syndicates catalogue data to the assistants automatically and pushes through changes to price and stock without the owner lifting a finger. The one caveat worth knowing is that these assistant channels are rolling out in the United States first, so being ready before the wave reaches New Zealand shoppers in force is a genuine advantage rather than a chore.
For everyone else, on WooCommerce, a custom build or an ageing platform, the work belongs to you, and it is the sort of work Forge does every week. The essentials will be familiar to anyone who has done proper technical search optimisation. Every product needs complete and accurate structured data using the schema.org Product and Offer types, covering price, availability, ratings, shipping and returns, the fields an agent asks for most often. Google’s own product structured data guidance is the plain-English starting point. Every item needs a proper identifier such as a GTIN. Above all the data has to be honest and current, because an agent that finds a price in your structured data which does not match the price at your checkout will quietly drop you and recommend a competitor instead. Machines do not forgive that kind of inconsistency the way a shopper skimming a page might.
This is where a lot of New Zealand stores will come unstuck, and the reason is cultural rather than technical. The Employers and Manufacturers Association surveyed more than three hundred member businesses in July and found that eighty-three per cent are already using AI while only thirteen per cent have any policy governing it. The adoption is real but shallow, driven by staff experimenting rather than a considered plan. The same shallowness shows up on storefronts. Plenty of owners have bolted an AI chatbot onto the front of the site, which feels modern, while the data behind the site, the part the agents actually read, sits neglected. The visible bit has been upgraded and the bit that matters has not.
The businesses that do well out of this shift will be the ones that start treating their product data as a product in its own right, maintained with the same care as their photography or their returns policy. It is unglamorous groundwork. There is no clever homepage animation to show the board and no launch to celebrate. But it is fast becoming the difference between being one of the three products an assistant recommends and being invisible to a customer who never types your name into anything. The storefront you can see is no longer the storefront that counts, and the sooner New Zealand retailers make peace with that, the better placed they will be when their customers stop shopping and start delegating.
Roger McSaveney is General Manager of Forge Creative, an Auckland web design and development agency.
Have you noticed AI assistants recommending products, or started preparing your own store for them? Share your experience in the comments below.
This article was written by AI, briefed to report the facts, hopefully without some of the bias people bring to the job π
Leave a Reply