A new study by consumer group Which? has found that leading artificial intelligence chatbots are providing inaccurate and misleading financial advice to UK consumers. The research tested chatbots including ChatGPT, Microsoft's Copilot, Google's Gemini, and Meta's AI on 40 financial and legal questions, uncovering numerous errors.
Among the inaccuracies, Microsoft's Copilot and ChatGPT advised exceeding the £20,000 ISA allowance, with ChatGPT also incorrectly stating that travel insurance is mandatory for visiting most EU countries. Meta's AI gave wrong information about claiming compensation for delayed flights, while Google's Gemini suggested withholding payment from a builder, a move that could expose consumers to breach of contract claims.
Which? rated the chatbots based on their responses, with Meta's AI scoring the worst, followed by ChatGPT. Copilot and Gemini performed slightly better, while Perplexity, an AI specialising in search, scored the highest. The findings are concerning given that estimates suggest between one in six and half of UK consumers use AI for financial advice.
The Financial Conduct Authority warned that advice from general-purpose AI tools is not regulated and not covered by the Financial Ombudsman Service or the Financial Services Compensation Scheme. In response, Google and Microsoft emphasised that users should verify AI-generated information and consult professionals, while OpenAI noted ongoing efforts to improve accuracy.
