Picture a couple sitting across the desk while a lending strategy gets laid out — borrowing capacity, a loan structure, maybe a plan to add an investment property. The numbers get read out, and one partner is nodding along, already three steps ahead. The other’s face does something quieter: a flicker of confusion, a shift in the chair, a look that says “I don’t actually follow this” or “I’m not comfortable with this” — but nothing gets said out loud. A skilled broker sees that look and stops to check in. An AI chatbot, generating the same numbers on a screen, sees nothing at all. It just keeps going.
That gap — between an answer that sounds right and advice that actually lands with both people in the room — is where a lot of Australians are about to get an expensive lesson in the difference.
How Many Australians Are Actually Using AI for Money Decisions
This isn’t a fringe behaviour anymore. Recent research from AustralianSuper found 17% of Australians are already using AI for financial decisions, rising to 28% of 18–29 year-olds, against just 6% of those over 60 (AustralianSuper, 2026). Compare Club’s AI Index puts current usage at a similar 16.6%, with a further 30.9% planning to use AI for financial guidance soon — meaning close to half the country is either already leaning on AI tools for money decisions or intends to (reported in Professional Planner, 2026). Globally, the picture is similar: Vanguard’s global chief economist, Joe Davis, has gone as far as calling ChatGPT “probably the largest provider of financial advice in the world” (ifa.com.au, 2026).
The trend is sharpest among younger Australians. A 2025 RACQ Bank survey of more than 500 young Queenslanders found 45% had followed AI-generated financial advice without consulting an expert, and nearly half said they were confident in its accuracy (Savings.com.au, 2026). Separately, research from the Council of Australian Life Insurers found 60% of Australians would rely on AI-generated financial advice for money decisions, while more than half said they’d actually prefer a professional if advice were easier to access and less costly (Insurance Business, 2026) — a reminder that the appeal of AI is often less about AI itself and more about advice feeling out of reach.
Trust hasn’t fully caught up to usage, though. AustralianSuper found 37% of Australians distrust AI for financial information, and of the people who do use it, only 54% cross-check what it tells them, while 25% have acted on AI-generated advice without verifying it at all. ASIC flagged the same trend in its Key Issues Outlook 2026, warning that “consumers face risks from automated decisions, AI-driven interactions, and scams amplified by technology” (ASIC, 2026).
Why So Many People Are Turning to AI First
None of this happens in a vacuum. For HENRY households — high income, high tax, high mortgage, and somehow still no clear sense of getting ahead — AI is filling a real gap: it’s free, it’s available at 11pm when the anxiety hits, and it never makes you feel silly for asking a basic question. It’s also filling a gap the industry itself has flagged — the Council of Australian Life Insurers research above found more than half of Australians would still prefer a professional, if only advice were easier to access and didn’t feel gatekept behind cost or a minimum portfolio size. A lot of people are also simply paralysed by conflicting advice from parents, colleagues, and finfluencers, so an AI chat that sounds calm and authoritative is genuinely appealing.
The problem isn’t that people are curious or want to educate themselves — that’s healthy. Even ASIC’s own Moneysmart service is upfront that AI has legitimate uses: it’s good at breaking down complex information, explaining basic concepts, and pointing you toward topics worth researching further — “a bit like a friendly commerce or business studies teacher,” as Moneysmart puts it. The problem is where the line gets crossed from researching with AI to deciding based on AI, often without noticing the line was crossed at all.
What AI Actually Gets Wrong
This is where the evidence gets uncomfortable. A September 2026 UK report by fintech research firm Saturn — Artificial Authority: Should You Trust AI to Deliver Financial Advice? — tested 18 popular AI models, including ChatGPT, Claude, Copilot, Grok, and Gemini, against 121 real financial questions repeated five times each for consistency: more than 10,000 responses in total. The overall error rate was 57%, climbing to 88% on complex questions; the worst-performing model got it wrong 82% of the time, and even the best-performing reasoning model still erred 39% of the time. In one worked example, an AI-generated pension answer left the user exposed to a £17,500 unnecessary tax bill that correct advice would have avoided (IFA Magazine, 2026).
A peer-reviewed study from the Technical University of Munich and the University of St Gallen ran a cleaner test: asking ChatGPT, Gemini, and Copilot to build investment portfolios for different ages and risk appetites. Every model’s recommendations skewed further from a sensible, diversified benchmark than they should have — more than 93% concentrated in US stocks against a 59% benchmark weighting, consistent overweighting of trendy sectors like technology, and a bias toward recently popular stocks. The researchers’ conclusion: large language models “exhibit similar cognitive biases as human investors” (Hartmann, Winder & Hildebrand, PLOS ONE, 2025).
ASIC’s own Moneysmart team has been blunt about the core mechanical risk: general-purpose AI can “hallucinate, or invent, information that has no basis in fact.” A financial adviser interviewed by the ABC put it more plainly — AI “often gets figures wrong, it makes sort of wrong assumptions” — and Philipp Winder, one of the University of St Gallen researchers behind the portfolio-bias study above, described large language models to the ABC as “black boxes” that “just try to predict the next word,” warning that trusting them with real investment or borrowing decisions “might be a bit dangerous” (ABC News, 2025).
Moneysmart’s own guidance is the clearest summary of the structural problem: general-purpose AI “doesn’t really know you.” It only knows what you happen to type into a chat box, it hasn’t been trained to ask the clarifying questions a real adviser would ask before forming a view, and — critically — it hasn’t been designed or licensed to give personal financial advice in Australia at all (Moneysmart, 2026).
That last point is the one worth sitting with, because it’s not a quality issue that better prompting fixes. It’s a licensing issue.
The Law Most People Don’t Know Exists — And Why It’s There
Every time you sit down with a mortgage broker or financial adviser in Australia, you’re dealing with someone operating inside a legal framework most clients never think to ask about. It’s worth understanding, because it’s the exact thing an AI chatbot doesn’t have.
The Australian Credit Licence (ACL) system. Under the *National Consumer Credit Protection Act 2009* (NCCP Act), anyone providing credit assistance — arranging a home loan, comparing lenders, advising on how to structure borrowing — must operate under an Australian Credit Licence issued by ASIC, or as a credit representative authorised under someone else’s licence. This isn’t paperwork for its own sake. It replaced a patchwork of inconsistent state-based rules with a single national regime specifically so consumers would have one clear, enforceable standard no matter which broker they sat down with, anywhere in the country.
The chain of accountability. A credit representative doesn’t operate in isolation — they sit under a licensee who is legally responsible for their conduct, who must maintain professional indemnity insurance, who must have a dispute resolution process, and who must report serious compliance breaches to ASIC. Build & Protect Financial Services, for example, operates as Credit Representative #539491 under AFG’s Australian Credit Licence #389087 — a structure that means there’s always a licensed entity accountable for the advice given, and always a path of recourse through the Australian Financial Complaints Authority (AFCA) if something goes wrong.
The best interests duty. Since 2020, mortgage brokers have carried a specific, non-negotiable legal duty under the NCCP Act: to act in the consumer’s best interests, and to prioritise those interests over the broker’s own, the lender’s, or anyone else’s when there’s a conflict. ASIC’s Regulatory Guide 273 sets out exactly what this looks like in practice — brokers must make reasonable inquiries into a client’s objectives, requirements, and financial situation, assess a range of suitable products against that specific situation, and be able to demonstrate why the recommendation actually serves the client rather than just disclosing a conflict and hoping that’s enough (ASIC RG 273). Disclosure alone was never going to cut it — the duty exists precisely because “we told you about the conflict” and “we acted in your interest” are two very different things.
This duty didn’t appear out of nowhere. It’s a direct response to the Banking Royal Commission, which uncovered exactly what happens when the person giving you financial guidance has an unspoken incentive to point you toward the option that suits them, not you. The best interests duty and the licensing regime that sits behind it exist for one purpose: to make sure the person across the table from you is legally required — not just professionally encouraged — to put your outcome ahead of theirs, and to be personally and financially accountable if they don’t.
Why None of That Applies to AI
Here’s the uncomfortable part for anyone typing a mortgage question into a chatbot: none of the above exists on the other side of that conversation.
A general-purpose AI tool has no Australian Credit Licence. It has no best interests duty. It makes no reasonable inquiries into your actual circumstances beyond whatever you happen to type — and Moneysmart is explicit that it hasn’t even been trained to know which clarifying questions matter. If it gets your situation badly wrong, there’s no AFCA complaint to lodge, no professional indemnity insurance standing behind the advice, and no licensee accountable for the outcome.
This isn’t a hypothetical gap. A 2026 UK Financial Conduct Authority survey of 666 young investors found most people using AI for money decisions don’t actually understand that this gap exists: 44% incorrectly believed AI-generated investment information was regulated, and 32% wrongly thought they’d be compensated through a financial ombudsman scheme if that advice turned out to be wrong (FCA, 2026). The FCA’s own director of consumer investments, Lucy Castledine, was careful to frame where AI actually fits: it “can help you research companies, understand jargon or explore options before you make a decision” — not make the decision itself. The same gap applies here — there’s no AFCA complaint to lodge against a chatbot, because a chatbot isn’t a credit licensee or a financial services licensee in the first place, in Australia or anywhere else.
That’s not a criticism of the technology — it’s simply not what it was built to do. The entire licensing and accountability framework in Australian credit and financial advice exists to solve a problem AI hasn’t been designed to solve at all: making sure someone is legally on the hook for getting your specific situation right.
The One Thing AI Will Never Be Able to Do: Read the Room
This is the part of the conversation that rarely gets discussed, and it matters more than most people realise — especially for couples making a joint decision.
A genuine best interests inquiry, done properly, isn’t a form. It’s a conversation where a broker is watching for the things that don’t get typed into a chat box: the hesitation before someone answers a question about risk, the partner who keeps deferring to the other on money, the couple who both said “we’re aligned” in the first five minutes but whose body language told a completely different story by minute twenty. Under RG 273, a broker assessing a joint loan is required to make reasonable inquiries into the objectives and circumstances of both borrowers — not just whoever happened to be typing.
That matters more than it sounds like it should, because Australian couples are far less financially aligned than most people assume. Finder’s 2025 survey of partnered Australians found that 41% — roughly 4.8 million people — don’t fully combine their finances, and only 59% share everything equally. The generational gap is stark too: 77% of Baby Boomer couples share finances fully, compared with just 31% of Gen Z couples, many of whom deliberately hedge with a mix of joint and separate accounts (Finder, 2025). Money expert Rebecca Pike’s advice to couples in that same research is telling: successful joint financial decisions require “honesty — there should be no financial secrets or unexpected surprises, which can lead to conflict.”
An AI chatbot has no way to surface any of that. It answers the question it’s given, from the person who typed it, based on the assumptions embedded in the prompt — it has no way to notice that “we want to buy an investment property” actually means “I want to buy an investment property and I’m hoping the numbers will help me convince my partner,” or that one partner’s discomfort with a strategy is being talked over by the other’s enthusiasm. It cannot ask the follow-up question that a real adviser asks when something doesn’t add up, and it cannot pick up on the pause that tells you the real conversation hasn’t happened yet.
This is, genuinely, one of the most valuable — and least visible — parts of what a good broker or adviser actually does: not just running the numbers, but making sure both people in a joint decision are actually making it together, with the same information and the same understanding of what they’re signing up for. No amount of AI sophistication replaces a human noticing that the room isn’t aligned.
How to Use AI Sensibly Without Letting It Make the Decision
None of this means AI is useless — used well, it’s a genuinely good research tool. The distinction that matters is between using it to get smarter before a real conversation, versus using it instead of one.
| Good use of AI | Where it needs a human instead |
|---|---|
| Understanding what a term like “offset account” or “LVR” means | Deciding which loan structure actually suits your situation |
| Getting a rough, general sense of how borrowing capacity works | Getting an actual number based on your real income, debts, and goals |
| Preparing questions to bring to a broker or adviser | Answering questions about your specific circumstances |
| Comparing broad pros and cons of strategies you’ve heard of | Recommending a strategy and being accountable if it’s wrong |
| Researching independently before a joint conversation with your partner | Replacing the joint conversation itself |
A few practical habits worth adopting: treat AI output the way you’d treat a first-year opinion, not an expert one — verify anything material against a primary source or a licensed professional before acting; never paste sensitive financial details, account numbers, or income documents into a general-purpose AI tool, since you have no visibility into how that data is stored or used; and if you’re making a decision as a couple, do the AI research separately, then have the actual conversation with each other — and with your broker — out loud.
If your AI research has thrown up questions about how your existing loans are structured — or whether your debt is actually working as hard as it could be — our guide to deductible vs non-deductible debt is a good next read before that conversation.
Common Mistakes We See
- Treating a confident tone as a sign of accuracy. AI is optimised to sound certain, not to be correct — the two are not the same thing.
- Only one partner doing the research. A joint financial decision that’s only been “checked” by one person typing questions into a chatbot isn’t actually a joint decision yet.
- Skipping the accountability question. Before acting on any piece of financial guidance, it’s worth asking: if this is wrong, who is responsible, and what recourse do I have? For a licensed broker or adviser, there’s a clear answer. For a chatbot, there isn’t one.
- Using AI as the final word on borrowing capacity. Real serviceability depends on lender-specific policy, your actual liabilities, and how your income is assessed — figures that shift constantly and that a generic model has no live access to.
Frequently Asked Questions
Is it illegal to use AI like ChatGPT for financial advice in Australia?
No — using AI to research or understand general concepts isn’t illegal. What’s not happening is any licensing, accountability, or best interests obligation on the AI’s part, unlike a broker or adviser operating under an ACL or AFSL.
What’s the difference between a credit representative and a broker’s licensee?
A credit representative is authorised to provide credit assistance on behalf of a licensee, who holds the actual Australian Credit Licence and carries legal responsibility for the representative’s conduct, compliance, and professional indemnity cover.
What is the best interests duty and when does it apply?
It’s a legal obligation under the NCCP Act requiring mortgage brokers to act in the client’s best interests and prioritise those interests over their own or a lender’s when a conflict exists. It’s applied to every piece of credit assistance a broker provides, including reasonable inquiries into a client’s full circumstances — both borrowers’, in a joint application.
Can AI replace a mortgage broker?
It can replace some early-stage research and general education, but it can’t replace the licensed, accountable, personalised assessment a broker is legally required to provide — and it can’t replace a human noticing when two people making a decision together aren’t actually on the same page.
What should I do if I’ve already made a financial decision based on AI advice?
Bring what you found to a licensed broker or adviser for a proper review before acting further. It’s far cheaper to check the numbers now than to unwind a poorly structured loan or investment later.
The Real Question Isn’t AI vs Human — It’s Who’s Accountable
AI isn’t going anywhere, and for research and general understanding, it’s a genuinely useful tool to have in your corner. But there’s a reason the credit and financial advice industries in Australia are built around licences, representatives, and a legally enforceable duty to act in your best interests: someone has to be accountable for getting your specific situation right, and someone has to be capable of noticing when the human side of the decision — two people, two sets of expectations, one big commitment — isn’t as aligned as it looks on the surface.
If you and your partner are weighing a borrowing decision, a refinance, or an investment strategy and want more than a confident-sounding answer, book a strategy session with the team at Build & Protect Financial Services. We’ll look at your actual numbers, ask the questions an AI never will, and make sure whatever you decide, you’re deciding it together.
This article is general information only and doesn’t take into account your personal financial situation. It isn’t tax or financial advice — speak with a licensed adviser or broker about how this applies to you. Build & Protect Financial Services Pty Ltd, Credit Representative #539491, is authorised under Australian Credit Licence #389087 (AFG).