A Manhattan Associates survey landed a genuinely strange result this year: 47% of Australian shoppers now say they trust AI recommendations more than a human sales assistant.

Not slightly more. More. That’s according to reporting from eCommerce News Australia, and it tracks with what most of us have quietly noticed ourselves. The algorithm knows your size. It remembers you hate polyester. It doesn’t try to upsell you on a warranty you don’t need.

But here’s the part that doesn’t get nearly enough attention. Trusting AI to suggest a jacket is not the same as trusting it to move your money.

Ask the same shoppers whether they’d hand an AI system unsupervised control of their bank account, their crypto wallet, or their next big purchase, and the number falls off a cliff. We trust the recommendation.

We don’t trust the transaction. That gap is the real story, and it says more about how Australians think about risk than any single survey stat ever could.

The Recommendation Is Cheap. The Transaction Isn’t.

Think about why an AI recommendation feels low-stakes. If the algorithm gets it wrong, you return the jumper. Mild annoyance, fifteen minutes at Australia Post, done. Nobody loses sleep over a bad product suggestion.

Money is different. A wrong recommendation costs you a returns label. A mishandled transaction costs you actual funds, and possibly your data along with them. PayPal’s Australian research found that even among shoppers who already use AI tools to browse and compare, a much smaller slice are comfortable letting AI actually execute a purchase without a manual confirmation step.

The threshold for autonomy drops sharply the moment real dollars are on the line, and that’s not irrational. It’s exactly the risk calculus you’d want people making.

This is where the conversation gets interesting for anyone who has watched other high-stakes digital sectors solve the same problem. Verified payout systems, licensing transparency, and clear identity checks are precisely how industries built entirely around moving other people’s money have tried to close this exact trust gap.

Online gambling platforms are one of the clearest examples, because unlike a retail app, the entire product depends on players believing their money will actually come back out the way it went in.

Look at how operators earn that confidence, and an overview of the best online casinos in Australia tends to rank platforms less on flashy AI-driven personalisation and more on the boring stuff: licensing, payout speed, and whether the terms match what was advertised.

Gambling carries real financial risk and isn’t for everyone. If it stops being fun, Gambling Help Online and 1-800-GAMBLER are there.

That’s the pattern worth noticing. Trust with money is earned through verification, not polish.

What Actually Closes the Trust Gap (Hint: It’s Not a Better Chatbot)

Here’s a claim that will annoy some product teams: throwing a slicker AI interface at a payment flow does almost nothing to fix consumer hesitation. It’s not a UX problem. It’s a proof problem.

Consider what KPMG found in its 2024 Generative AI Consumer Trust Survey. People weren’t asking for AI to be smarter. They wanted to know who was accountable when it got something wrong, and what recourse they had.

That’s a governance question dressed up as a technology question. Nobody solves it by adding more machine learning to the checkout page.

Boston Consulting Group projects consumer trust in AI will climb by roughly 15 percentage points by 2030, according to BCG’s recent analysis. But that curve isn’t flat across use cases.

Trust in AI shopping suggestions is already climbing fast. Trust in AI handling unsupervised financial decisions is climbing much slower, and for good reason. One mistake is annoying. The other is expensive.

Three things reliably move the needle, and none of them involve a chatbot:

• Clear, published terms that don’t require a law degree to parse
• Fast, verifiable settlement, so users see proof the system works rather than a promise that it will
• Human recourse when something breaks, not a support bot that loops the same three canned replies

Financial advisors are running into the same wall. A FINRA study covered by Financial Planning found people trust AI roughly as much as a human advisor for basic questions, but that trust narrows fast once the advice starts touching actual portfolio decisions. Same pattern. Same wallet effect.

Generational Splits Are Bigger Than Most Reports Admit

Not every Australian is walking into this trust gap the same way, and lumping them together flattens something worth seeing clearly.

Younger shoppers are noticeably more comfortable letting AI drive the whole process, recommendation through to payment.

EMarketer’s research on AI shopping trust found Gen Z consumers, across several markets including Australia, are markedly more willing to let an algorithm make the final call on a purchase than older cohorts are.

Some of that is comfort with the tech. Some of it is just having grown up watching Amazon get delivery estimates right often enough to stop double-checking.

Older Australians tend to hold the line further out. They’ll take the AI’s suggestion. They still want a human, or at minimum a very obviously transparent process, standing between the suggestion and the money actually moving.

That’s not technophobia. Plenty of them use internet banking without blinking. It’s a specific, learned caution around automated systems making unsupervised financial calls, and honestly, it’s aged well given how many “smart” systems have quietly mishandled subscriptions and auto-renewals over the past few years.

The businesses getting this right aren’t the ones with the most advanced AI. They’re the ones being explicit about where the AI’s job ends and a verifiable, auditable process begins.

Where This Actually Lands for Australian Businesses

If you’re building or buying into an AI-driven commerce tool, the lesson from this data isn’t “make the AI better.” It’s “make the money part boring, transparent, and provable.” That’s unglamorous advice, but it’s the advice that survives contact with actual consumer behaviour.

Retailers pushing AI further into the funnel, past the recommendation and into the checkout, are going to hit exactly this wall unless they pair it with radical transparency about how the transaction is handled and who’s accountable if it fails.

The 47% trust figure is a genuine opportunity. It’s also a ceiling, not a floor, until the money side catches up to the advice side.

Australia’s broader digital trust posture has been a live conversation for a while now, and this fits squarely inside it. We adopt fast. We verify slower. Any business, or algorithm, hoping to close that gap needs to earn the second part, not just impress with the first.

Frequently Asked Questions

• Why do Australians trust AI for shopping advice but not for payments? Recommendations carry low stakes. A bad suggestion costs a few minutes and a return. A mishandled payment costs actual money and personal data. Consumers apply a much higher verification bar once real financial risk enters the picture, which is a rational response, not a technology hang-up.
  
  
• Does trust in AI vary by age group in Australia? Yes, noticeably. Younger shoppers, particularly Gen Z, show more comfort letting AI drive purchases end to end. Older Australians are more likely to want a visible, verifiable human or process step between an AI suggestion and the actual transaction.
  
  
• What actually builds consumer trust in automated financial systems? Clear published terms, verifiable and fast settlement, and genuine human recourse when something goes wrong. Polished interfaces and smarter algorithms rank far lower in surveys than transparency and accountability when consumers are asked what earns their trust.
  
  
• Will AI trust in financial decisions keep growing? Industry forecasts suggest yes, gradually. BCG projects a meaningful rise in consumer trust in AI by 2030, but growth is uneven. Low-stakes use cases like shopping recommendations are climbing fastest, while high-stakes financial decisions remain far more cautious