New Zealand small businesses are embracing AI fast. But new research suggests many are preparing for only half the shift, and the other half is harder to measure than it looks.
The Prospa SME Sentiment Tracker, run by Clarity Insight surveying 501 SME decision-makers in late August and early September, found 59 percent now use AI in some form, up from 50 percent in March. The share using no AI at all fell from 45 percent to 36 percent.
For the hold-outs, cost isn’t the main barrier. Uncertainty about return on investment (34 percent) and not knowing which tools suit their needs (32 percent) top the list. Upfront cost ranks sixth, at 12 percent.
“The hold-outs are shrinking fast. What’s stopping the rest isn’t the price tag. It’s that nobody has shown them which tool actually solves a problem they have,” says Adrienne Begbie, Prospa’s New Zealand managing director.
The other side of the counter
That’s AI inside the business. Customers are adopting it too, and changing how they find businesses.
PwC research, based on an Ipsos survey of 1,000 New Zealanders who had used AI for shopping, found 44 percent had cut back on search engines when researching purchases. Most use AI to compare products and prices (64 percent) or research planned purchases (62 percent).
Jono Swanston, General Manager of Sprocket Digital, says businesses where customers research and compare before buying are most exposed, citing “professional services, tourism, retail, e-commerce and specialist manufacturers”.
“The concern is that businesses may not realise they’re losing visibility because they’re still looking at their traditional search rankings and assuming everything is fine.”
A study by Auckland-based performance marketing agency The Optimisers, The NZ AI Answers Project, suggests visibility is harder to pin down than many assume.
In July, 76 New Zealanders asked ChatGPT, Google AI Mode and Microsoft Copilot the same three questions: The best KiwiSaver provider; the best power company, and; the best small business accounting software. The engines disagreed on which provider to name first in 134 of 211 matched cases. Even within ChatGPT, 50 of 72 participants got a different top answer at least once when they switched off its memory.
Accounting software was the exception. All 221 accounting answers named Xero first – because, Richard Conway, CEO of The Optimisers says, every source agreed. Power had no consistent winner because the sources disagree.
He expects most SME categories to behave like power. “Most small business categories look more like power than accounting, and some will be worse. If you ask who the best commercial electrician in Hamilton is, there is no Canstar Award and no Consumer NZ survey to anchor the answer, so the engines fall back on directories, review sites and listicles.”
Copilot leaned most on SEO listicles, he says, and was the only engine caught citing AI-generated content farms – while giving the most confident answers. “Thin sources, confident answers. The upside is that an unanchored category is an earnable one.”
Richard Conway is a thought leader on AI in New Zealand, and regularly presents on the topic at industry events.
The cost to earn
Sprocket estimates Kiwi businesses could collectively spend more than $300 million over five years adapting to AI search. Swanston says the figure assumes 10 percent of the 17,280 NZ businesses with 20 or more staff spend an average of $3,000 a month for five years.
“It’s an indicative estimate rather than a prediction of exactly what businesses will spend, and we’d expect some of that money to come from existing SEO and digital marketing budgets rather than all being additional expenditure.”
Conway is however more cautious. “I am wary of big round figures in this space generally,” he says. Although, though he adds “the direction is right and the money is smaller and more specific than people fear”.
Both agree a single search tells you little. “You can ask ChatGPT the same question twice and get slightly different answers, let alone compare what it tells you with Gemini or Claude,” Swanston says.
But they differ on how to check. Swanston suggests 10 to 15 customer-style questions across ChatGPT, Gemini and Claude, repeated weekly for three or four weeks. Conway favours depth: Ask ChatGPT, Google AI Mode and Copilot your customer’s question ten times each in a fresh chat, noting who’s named first, which sources are cited and how your business is described – then get a colleague or customer to repeat it on their own account. “They can do it for free in a morning,” he says.
His key rule: “Count recommendations, not mentions.” In the study, Electric Kiwi appeared in 214 of 221 power answers but was named first in only 55. “If you only count mentions you will think you are winning when this is not actually the case.”
They split on paid tracking tools too. Swanston says specialist tools can automate the analysis and show why competitors appear ahead of you. Conway, asked whether SMEs should pay for them: “Not yet, for most of them, and I say that as someone who uses the tools every day.”
And on where the work happens. Swanston focuses on the business’s own site – not usually a rebuild, but “improving the information that’s already there and making sure it’s easy to find and understand”, with content that is “genuinely helpful rather than just another sales pitch”.
Conway argues “most of the battleground is not your website”. His advice: “Spend the money on the sources instead: The comparison sites, review platforms and industry bodies the engines use to formulate their recommendations.”
Don’t panic, but don’t wait
Neither suggests abandoning Google, and both say SMEs can start for free.
“I think it’s a now problem, but that doesn’t mean every small business needs to rush out and spend thousands of dollars tomorrow,” Swanston says.
“I wouldn’t recommend spending a large amount of money until you’ve worked out whether you’ve got a problem in the first place.”
Conway puts AI search “inside existing search budgets”, and suggests many owners are already getting AI-driven visits without realising, because the traffic often shows up in analytics as branded search. “Most owners are already getting it and crediting it to something else.”
His order for time-poor owners: Get the fundamentals right – Google Business Profile, reviews and a website that says plainly what you do and where. Then run the morning check. Then fix the one source that’s most wrong about you.
Jono Swanston is General Manager of Sprocket Digital.
Beyond the first check
You’ve asked the AI engines about your business. Here’s what the experts suggest doing next.
- Ask like a customer: “Someone asking AI to recommend a reliable plumber for an urgent hot water repair in Auckland is a much more useful test than asking what it knows about your business,” says Swanston.
- Watch the wording: Note the adjectives attached to your name each time it appears, not just whether you’re named. In The Optimisers’ study, Xero was almost always introduced with the same description, which shows how AI settles on a fixed line about a brand.
- Fix your own site first. “It should be clear about your areas of expertise, who your customers are and what makes you different, but it’s also worth looking at what other websites say about you,” Swanston says.
- Then spend in this order: Conway’s priority list:
• The directories, review platforms, industry awards and comparison pages in your category.
• Your own comparison and “alternatives” pages, both of which earned citations in the study.
• A regular check. “Three mornings a year is enough to see movement.” - Know when a tool is worth it: “Pay for a tool only when you have a contested category, a decent budget and someone whose job it is to act on the output,” Conway says.
