AI has made financial research faster, more thorough, and more accessible than it has ever been.
Ask it to explain what an endowment plan is, compare term versus whole life insurance, or how miles credit cards work—it handles all of this faster and more comprehensively than any human starting from scratch. In Singapore, 81% of adults are already using it to manage their money. More than half for personalised financial advice.
Across Asia, the picture is starker. In a single year, the share of households using generative AI for financial advice jumped from 18% to 60%. In that same period, the share of households considered financially secure fell from 32% to 25%.
Two numbers. Same study. Same year.
Greater access to financial information has not consistently translated into stronger financial confidence or preparedness. That correlation is our starting point. AI has made people more active in their financial decisions. It hasn't made those decisions better. Here's why.
What AI genuinely does well
Our credit card comparison page alone averages over 100,000 sessions a month. AI has made this kind of research even easier. Ask it to surface the top cashback cards in Singapore, rank personal loan rates by EIR, or explain how a balance transfer works—it handles all of this faster and more comprehensively than any human starting from scratch.
For anyone trying to understand what financial products exist, it compresses hours of research into minutes.
The problem isn't the output. It's what happens when someone treats a comparison as a recommendation.
But a recommendation requires something a comparison doesn't
When you ask an AI tool to recommend a credit card, it works without information that determines whether that recommendation is right for you specifically.
Your CBS credit score. Credit Bureau Singapore issues scores on a 1,000–2,000 scale. That number determines whether you'll be approved for the card being recommended. AI has no access to it. The best card in Singapore is irrelevant if you don't qualify for it.
Your existing card portfolio. If you already hold cards with overlapping cashback categories, an additional card in the same tier duplicates rather than expands your setup. AI doesn't know what's already in your wallet, so it can't tell you whether a new card adds value or just replicates what you have.
Your actual spend pattern. Not what you assume you spend on dining or groceries. What your statements show. A card earning 8% on dining delivers that rate only on dining transactions. If dining is 15% of your monthly spend, the remaining 85% earns at base rate. AI recommends on category; it doesn't know your category mix.
Your income relative to minimum spend. Many of the most rewarding cards in Singapore require $500 to $1,000 in monthly spend to unlock the advertised rate. AI has no way of knowing whether you'll hit that consistently without changing your behaviour, so it assumes you will.
Without these four inputs, the output is generic by design. Not by failure.
Technically, you could feed AI these details…
Enter your spend categories, monthly amounts, existing cards, and income range into a well-prompted AI and you'll get a more useful answer than a generic query returns.
But most people don't know that's what determines whether a recommendation is right for them. They ask "what's the best cashback card in Singapore" without knowing that spend category volume, portfolio overlap, and minimum spend tolerance are the variables that change the answer entirely. If you don't know what context to provide, you won't provide it—and the output defaults to generic.
We could actually see this behaviour manifest through the card matching tool on our site. ~31% of people who clicked to start didn't finish—dropping off before answering all the required inputs, namely reward preference, spend category, and amount spent. The questions used to determine whether a recommendation is right for you are the same ones people find hardest to answer.
There's also a privacy dimension most people miss. Free-tier AI tools use conversation data to improve their systems unless you actively opt out, meaning sensitive financial details shared in a query may not stay private.
And the more fluent you are with AI, the less likely you are to notice
Research published in the International Journal of Financial Studies found that overconfident users of AI financial tools are less likely to question the advice they receive—even when that advice is wrong. The study was specifically examining AI use in the context of financial fraud, but the observed behaviour is telling: the more fluent someone is with AI, the more they trust the output.
MAS recognised this dynamic early. In November 2025, MAS released a consultation paper on AI Risk Management in financial services—setting out regulatory expectations for how financial institutions should govern AI use.
The concern wasn't that AI produces bad comparisons. It's that comparison is not the same as advice, and the gap between the two has real consequences when people treat one as the other.
The adoption curve is steep. Awareness of this gap is not.
Use AI. Just don't confuse it for advice.
The quality of any financial recommendation is only as good as the inputs behind it. AI is a research tool—exceptionally good at telling you what exists. What it can't do is tell you what fits, because fitting requires knowing you.
The inputs that matter most are the ones most people never think to provide.
Curious what cards actually fit your spend pattern? Our card matching quiz starts with your situation, not the product.