8 money problems you should not ask ChatGPT for including 'serious' mistake

It has fast become the everyday sounding board for millions of Brits seeking advice.

But using ChatGPT to solve your money problems can be risky and have serious consequences.

This week, Garfield AI recovered £7,000 of debt for a customer who paid just £7.50. It has been welcomed as good news because it is run by professionals and approved by the Solicitors Regulation Authority (SRA).

But turning to the AI tool to solve your personal money problems could cause you to lose money, not make it. These experts explain the money problems customers should swerve when using the tool.

Pensions

Colette Mason, Founder & AI Systems Architect at London-based Clever Clogs AI, said: "The success of Garfield AI is an example of a highly-regulated, rules-based legal scalpel doing its job. The danger is the reckless analogy that this justifies trusting a general-purpose LLM, a digital sledgehammer, with your pension.

"Financial advice is a context problem, not a maths problem. Untrained AI doesn't know you have a mortgage, debts, investments, or a low-risk tolerance. Its over-confidence creates a catastrophic psychological trap for novice users. MIT's Andrew Lo said the capability for LLMs to make real decisions is still 5 years away, with fiduciary duty being the biggest challenge. For now, AI should automate chores, not relationships, (e.g., portfolio balancing and capital loss advice). 

"Relying on general AI’s unverified output is a risk even for a trained advisor. For an untrained individual, someone not fluent in financial literacy, it’s a recipe for disaster. Until an AI assumes full fiduciary duty, it remains a smart toy, not a trustworthy advisor."

Tax, mortgages and insurance

Tony Redondo, Founder at Newquay-based Cosmos Currency Exchange commented: “AI excels at the heavy lifting of general research: understanding financial concepts and jargon, drafting complaint letters, organising budgets, and learning about financial products. 

"The danger emerges when clients use AI for specialist advice on irreversible, highly personalised matters like tax, mortgages, pensions, or insurance – where AI hallucinations could have serious consequences – rather than seeking professional help.”

Treating it like a confidant

Mitali Deypurkaystha, Founder/CEO at Newcastle upon Tyne-based Impact Icon AI commented: "There’s a fine line between using ChatGPT well and misusing it, and the consequences on either side are starkly different. ChatGPT is extraordinary at mimicking sentience, but that’s all it is: mimicry. 

"It has dazzling intelligence, but no wisdom, which comes from lived experience. Remember this and it becomes a powerful practice partner for interviews, presentations or difficult conversations. Its lifelike responses help you rehearse when you’re nervous or anxious, and test approaches for happiest outcomes. 

"It enhances your communication skills, attracting deeper human relationships. But cross that line, start treating it as a sentient friend, therapist or confidant, and the opposite happens. 

"Communication skills diminish, and isolation creeps in. In the UK we’ve already seen reports of psychosis, people falling in love with chatbots, and even a tragic suicide linked to someone mistaking ChatGPT’s mimicry for genuine sentience. Always use ChatGPT as a tool, not a human substitute."
 

Debt, products and tax 

Mo Saleh, CEO at Quantum Placements commented: "The Garfield AI case is a great example of technology creating real value, turning a complex and often intimidating process into something simple and affordable. 

"That is a genuine win for customers. With tools like ChatGPT, though, there are some clear lines to draw. Safe uses include summarising long financial documents, explaining jargon in plain English, drafting letters to creditors, or even role-playing negotiation scenarios before a real conversation. 

"These uses save time, reduce stress, and build confidence. Risky uses include relying on AI to calculate tax liabilities, restructure debts, or recommend financial products. 

“The danger here is that outputs can sound convincing but still be wrong, incomplete, or outdated. The risk becomes real when people move from using AI as a support tool to treating it as a substitute for regulated advice. Debt restructuring, pensions, investments, and tax planning all require context, regulation, and accountability.”