Using AI for financial advice? Experts urge caution

TL;DR:

• Some 28 million UK adults used AI for personal finance advice in 2025, according to Lloyds Banking Group.
• Experts warn the tools produce plausible but materially wrong answers, especially on tax and pension allowances.
• Unlike regulated advisers, AI has no fiduciary duty and offers little recourse when its guidance proves flawed.

AI chatbots have become a default first port of call for money questions, but specialists warn that treating them as authorities rather than tools invites expensive mistakes. With an estimated 28 million UK adults turning to AI for personal finance advice last year, the stakes of those errors are widening — and uptake is heavily skewed toward younger users, with 36% of 18-to-34-year-olds using the technology to inform investment choices.

Confident, plausible, and sometimes wrong

The core problem is that AI generates fluent language from patterns in its training data rather than reasoning the way a human adviser does. David Horowitz of accountancy firm Gerald Edelman says he has seen platforms miscalculate pension contribution headroom — ignoring tapering for higher earners or assuming unused allowances can be carried forward incorrectly — producing figures that “look entirely reasonable” but could trigger unexpected tax charges. Danilo McGarry, an AI adviser to the Chartered Institute of Personnel and Development, recalls a user asking whether to rebalance a portfolio without mentioning their time horizon or that holdings sat in a pension wrapper; the chatbot’s advice to sell would have caused an avoidable tax event.

The practical guidance is to be specific, cross-check answers across multiple models, and treat agreement between them as a weak signal of reliability rather than proof. A Which? study ranked Perplexity and Google’s Gemini highest for accuracy and responsibility, though experts suspect such rankings date quickly. Crucially, most mainstream platforms are not regulated financial advisers under UK law, carry no fiduciary duty, and offer limited recourse when wrong — which is why the Financial Conduct Authority stresses that consumers should check trusted sources before acting.

Looking forward

The regulatory picture is still forming. The Mills Review is examining AI’s medium-to-long-term impact on retail financial services and will report to the FCA board this summer. Until clearer guardrails exist, the burden sits with users — including a warning to guard personal data, since conversations may be retained and used to train future models. The pattern is familiar across UK financial services, where the FCA has cautioned that AI also makes financial crime cheaper and faster: the technology’s accessibility is precisely what makes its blind spots dangerous.