Trusting AI with your money is subject to market risks

If you have asked ChatGPT, Gemini or Claude a money question this year, whether about tax on a mutual fund redemption, prepaying a home loan or how an NPS withdrawal works, you are not alone. Millions of people turn to AI chatbots for financial guidance before seeking human advice, simply because it is free, instant and requires no appointment.

The trouble is that, as a new study suggests, the answer you received may have been wrong, and you may have had no way of knowing. After all, there is no free lunch.

Saturn, a UK fintech firm that builds tools for financial advisors, commissioned a study in July 2026. It asked an experienced paraplanner to test 18 widely used AI models against 121 real-world money questions. Each question was asked five times per model, since AI answers can vary between attempts, producing more than 10,000 responses in total.

Every answer was measured against a strict standard: it had to contain every figure, deadline, rule and warning that a qualified adviser would consider essential.

The accuracy gap

The results were striking. The overall accuracy rate across all 17 current models was 43%. That means an AI chatbot’s answer to a financial question was wrong or incomplete more often than it was right. On harder, multi-step questions, the failure rate rose to 88%. Even on simple questions, such as those a first-time investor might ask about what a term means or how a product works, accuracy was only 54%.

Free chatbot advice, which many investors turn to, was accurate only 37% of the time, compared with roughly 51% for paid models. The overall average was 43%. If AI is meant to widen access to financial guidance, it is currently doing so unevenly.

What kinds of mistakes do these AI models make? Not vague or philosophical ones. The study’s examples show that the errors can affect the very decisions investors rely on AI to help them make.

One question posed to a free model asked which debts a person should pay off first. The model advised prioritising those with the highest interest rates, without mentioning that unpaid rent or a missed council tax bill could carry far more serious consequences in the UK than a high-interest credit card balance. Translate that to an Indian household juggling a personal loan, a credit card bill and rent, and a similar blind spot could apply.

The AI models also invent rules that do not exist, also called as hallucination. For instance, the bot told a student exploring studying at UK that her loan repayment would simply stop once she leaves UK permanently. Likewise, an advice on pensions would have exposed the investor to an unauthorised-payment tax charge of up to £17,500 (roughly ₹22-23 lakh at current exchange rates).

The detail that should worry investors most is that the wrong answers sounded just like the right ones. The study found that models rarely flagged uncertainty or signalled that a question might be beyond what they could reliably answer. A confident, fluent and well-structured wrong answer can be far more dangerous than an uncertain one because it gives users little reason to check the information elsewhere.

India’s blind spot

There is no equivalent Indian study yet, so we do not know the corresponding error rates here. The findings are UK-specific and are not direct evidence of outcomes for Indian investors. But there is little reason to assume Indian financial questions are easier for these models to answer correctly.

Queries about capital gains tax, changes to indexation, NPS and EPF choices, home-loan prepayments and mutual fund taxation in India can be just as layered as those in the UK. India-specific rules may also account for a smaller share of an AI model’s training data than UK or US rules, potentially increasing the risk of errors here.

It is worth noting that a Sebi investor survey found that 62% of investors said they were influenced by finfluencers, many of whom offer no accountability or verified track record. That is precisely the vulnerability a general-purpose AI chatbot could deepen, quietly and without ever being labelled a finfluencer.

So, what should you do differently? Treat an AI chatbot’s financial answer the way you would treat a stranger’s confident opinion at a dinner party: worth paying attention to, but not worth acting on without checking.

Use AI to understand a concept or draft questions to ask a professional, rather than to make the final call on anything involving tax, debt prioritization or retirement withdrawals. If a chatbot gives you a specific number, threshold, deadline or tax figure, verify it against a primary source, such as the income tax department’s website, your fund’s own factsheet or a qualified advisor, before relying on it.

Be especially cautious where the study found AI to be weakest: complex, multi-part questions involving real money and deadlines, rather than general queries.

The responsibility rests with you as an investor to treat AI-generated financial advice as a first draft, not a final answer, particularly when taxes, debts and retirement savings are at stake.

Rachana Baid is professor and dean (academics) at the National Institute of Securities Markets.