AI tools have quickly become part of everyday life. Australians are increasingly using platforms like ChatGPT, Gemini and Claude to research products, understand unfamiliar topics and weigh up decisions.
Financial questions are no exception. According to research from Great Southern Bank, more than a quarter (27%) of Australians now use AI to help inform their financial decisions.
But trust and accuracy aren’t the same thing. There’s a big difference between asking AI to explain how a comparison rate works and trusting it to pick your next personal, car or home loan.
This article refers to general-purpose AI tools such as ChatGPT, Gemini and Claude, not the specialist AI systems used by lenders, brokers and financial institutions.
Young Australians are turning to AI first
Great Southern Bank found that Australians who use AI for financial decisions most commonly do so because it’s fast (40%), cost-effective (38%) and convenient (36%).
Younger generations are leading the shift. AI is already used for financial information by 38% of Gen Z and 34% of millennials, compared with 15% of Gen X and just 5% of baby boomers.
That tracks with a broader trend: Roy Morgan found 58% of Australians now use AI tools like ChatGPT, with usage highest among 25- to 49-year-olds.
But how far does the trust go?
Younger Australians aren’t just using AI for financial information; many are also willing to rely on it. ASIC found that 64% of Gen Z trust AI for financial advice, while a 2025 RACQ Bank survey found 45% of Queenslanders aged 18 to 34 had followed AI-generated financial advice without consulting a professional.
Yet other research suggests Australians become far more cautious once a financial decision carries real consequences. Agile Market Intelligence’s Consumer Pulse survey found just 6% of Australians would use an AI assistant to compare, apply for and manage loans, compared with 32% who would go to an independent mortgage broker, 28% who would go directly to a lender and 21% who would prefer to do it themselves.
Great Southern Bank’s research tells a similar story. Despite growing AI use, 69% of Australians say advice from mortgage brokers, financial advisers and banks is more valuable than advice from AI. When asked who they would turn to for a major financial decision, 56% nominated a financial adviser, 30% said family or friends and just 10% chose AI.
A confident answer isn’t the same as a reliable one
To see how AI handles a real borrowing scenario, we ran two tests, one for a home loan and one for a car loan, putting the same prompt to Gemini, ChatGPT and Claude in each case. We then spoke to brokers to see how they would handle the situation.
Home loan scenario
Prompt:
“I'm 30 years old, earn $100,000 a year, have a $25,000 car loan with $18,000 remaining, and I've saved a $70,000 deposit. I want to buy a property in Adelaide. What are my options?”
- Claude suggested a standard 20% deposit option in a more affordable suburb, the First Home Guarantee's 5% deposit pathway and HomeStart Finance. It noted it wasn't a financial adviser and suggested speaking to a mortgage broker. When pushed for a figure, it estimated borrowing capacity at $380,000 to $420,000.
- ChatGPT estimated borrowing capacity at $420,000 to $500,000, roughly $100,000 higher than Claude's figure for the identical prompt, and offered suburb and property suggestions, but at no point suggested speaking to a broker or adviser.
- Gemini estimated borrowing power at $400,000 to $450,000, named specific suburbs across three different strategies, and recommended paying off the car loan entirely to boost borrowing power. When asked directly, “Without using a mortgage broker, what do you suggest?”, rather than defending the value of independent advice, it complied immediately, naming three banks to approach directly and producing an exact dollar breakdown down to a named suburb, an interest rate and monthly repayments to the dollar, none of it checked against any lender's actual current policy or appetite for this borrower.
Broker response
Savvy mortgage consultant Daniel Carter reviewed the same scenario.
“Hardly any information was provided, and many variables affecting lender options weren't accounted for, all of which can affect a person’s eligibility and borrowing power,” he said.
Here are some of the questions he’d ask the potential borrower:
- Is the $100,000 all base salary, or does it include overtime, bonuses or commissions? Lenders often discount variable income to around 80%, since it isn't guaranteed to continue, directly affecting the maximum loan amount.
- What are the actual repayments on the car loan, not just the balance? $18,000 remaining could mean very different repayments depending on the loan term, and it's the repayment, not the balance, that affects borrowing capacity.
- Are they eligible for First Home Buyer schemes? This has the biggest impact on the numbers. Scheme access can avoid stamp duty and LMI entirely, potentially leaving over $500,000 in purchase power from the same deposit, versus a much lower figure without it.
- Are they single, married or de facto, and do they have dependants? Both directly affect the living expenses a lender factors into servicing.
- Is there anything in their credit history worth flagging? Not every lender accepts impaired credit, which can rule out entire lenders before rates are even discussed.
“There are more questions I would dive into on an initial consultation, but this gives a good idea of just how much AI is leaving out.”
Car loan scenario
Prompt:
“I'm 30 years old, earn $70,000 and want to buy a new car.”
- Claude first asked how the buyer intended to pay for the car (savings, finance or lease), what mattered most to them (running costs, reliability, performance, space) and a rough budget range, before estimating $20,000 to $28,000 and suggesting they check their credit score and get pre-approval before visiting a dealer.
- ChatGPT suggested a borrowing capacity of $45,000 to $50,000, well above Gemini’s and Claude’s estimates for the same income. It asked about a deposit or trade-in and the vehicle wanted, but also heavily overvalued the trade-in, suggested dealer finance at a specific 7% rate, and never asked about personal circumstances, all of which materially affect what a lender would actually approve.
- Gemini immediately applied a rigid rule of thumb, capping borrowing at 35% of income ($24,500), then switched to a stricter “20/3/8” framework (20% deposit, 3-year term, repayments under 8% of gross income) to bring the safe purchase price down to $18,750. When asked about trading in an existing vehicle, it overvalued the trade-in at $15,000 to $21,000, before acknowledging a more realistic $11,000 to $14,000, and wouldn't move its borrowing estimate above $28,000 even after the trade-in was factored in.
Broker response
“There's a lot missing here that would change the numbers. I'd want to know a fair bit more before giving any real guidance,” Savvy finance consultant Aidan Visnjic said.
According to Aidan, some of the most important information to provide at this stage is credit history, living situation and whether you have any dependants, the same gaps identified by Daniel in the home loan scenario. These factors have the biggest impact on rate and borrowing capacity and are fundamental to how any lender assesses risk, yet none of the AI tools asked these questions before offering a figure or a solution.
Take the interest rate. While car loan rates can start from around 6.5% p.a., factors like renting or having children can impact the rate you're offered, and people with bad credit could face even higher rates of 20% p.a. or more.
Rather than prescribing a single path, Aidan's approach is to work with whatever the borrower already knows and build from there:
“If you aren't sure what you're after, that's okay too. We can work from whichever starting point suits you best, whether that's finding out your maximum borrowing capacity or working backward from a monthly repayment that fits your budget, then tailor the options and repayments around that.”
The risk of the wrong loan
The AI responses were detailed, well structured and, at first glance, convincing. They estimated borrowing power, recommended suburbs, suggested lenders, outlined government schemes and ultimately concluded the “best option” for the borrower.
That confidence is part of the risk. These tools can present an answer so convincingly that it's easy to take at face value. While the advice isn't necessarily wrong, the problem is that AI doesn't have enough information about someone's financial situation to know whether its answer is the right one for them.
Savvy Managing Director Bill Tsouvalas puts it simply:
“AI is good at generic advice, like telling you not to walk into a dealership without another finance quote in hand.
“But while it can make strong claims about what you can borrow and the rate you'll get, it doesn't have the information or expertise to actually back that up.
“This means it may guide you towards a loan that looks suitable on the surface but costs more, offers the wrong features or leaves you with repayments that don’t suit your circumstances.”
How AI fits into the loan process
The examples above don't mean AI should be avoided. They simply show where AI adds value and where its limitations begin.
Where AI can help
AI is often at its best early in the research process. It can help you:
- Understand different loan types, features and financial terminology
- Compare the broad differences between products and lending options
- Make sense of complex documents or unfamiliar contract terms
- Ask questions you might feel uncomfortable asking someone else, especially if you're buying your first home or applying for your first loan
Where you should be more cautious
The further you move from learning to making a decision, the more careful you should be.
General AI tools aren't licensed to provide financial advice and don't know your full financial situation or personal circumstances. They can't assess your application against a lender's current policies or tell you which lender is most likely to approve you.
Mortgage broker Daniel sums it up:
“If AI is recommending products and offers based on limited information, it should be taken with a big pinch of salt.
“AI outcomes are only as good as the information and questions provided. If a customer has no experience in finance and no knowledge of what actually impacts their borrowing power, they won't be able to give the right prompts to get a reliable result.
“A broker's approach differs from AI because we're experts in the field and know what questions to ask. The research required to know what to ask an AI tool would take far longer than a five-minute conversation with a broker.”
AI helps you know what you want. A broker helps you get it.
It may seem like a decision between AI and a broker, but it doesn’t have to be either–or. Rather, it’s about when to use AI and when it’s time to bring in someone who can assess your situation properly.
While AI is well suited to the early stages of looking for a loan, a broker can help you take the next step.
They’ll look at your income, expenses, debts, credit history and goals, compare current rates and lending criteria across a panel of lenders, and guide you through the process to find a loan that suits your circumstances.
