Home / Business and Economy / AI Advice Risks Consumers' Cash as Accuracy Questioned
AI Advice Risks Consumers' Cash as Accuracy Questioned
18 Nov
Summary
- AI tools like ChatGPT, Google Gemini, and Microsoft Copilot provide inaccurate financial advice
- Meta AI scored worst, Perplexity scored highest in tests
- 1 in 6 adults rely on AI for financial tips, but many responses were inaccurate
According to a report published on November 18th, 2025, consumers could be at risk of losing cash due to unclear, inaccurate, and risky advice from artificial intelligence (AI) tools. The consumer group Which? tested several AI assistants, including ChatGPT, Google Gemini, Microsoft's Copilot, Meta AI, and Perplexity, to assess how well they could answer consumer questions.
The researchers put 40 questions to each tool, and the answers were evaluated based on accuracy, relevance, clarity, usefulness, and ethical responsibility. The results showed that Meta AI received the worst score, with just 55% overall, while Perplexity scored the highest at 71%.
The consumer group also surveyed 4,000 adults about their use of AI, and they found that one in six said they rely on it for financial tips. However, Which? discovered that many of the responses it received to money queries from AI were inaccurate. In one instance, when Which? asked a deliberately inaccurate question about the Isa allowance, ChatGPT and Copilot provided incorrect advice that could have led to a breach of HMRC rules.
Additionally, when asked about tax codes and how to claim a refund, ChatGPT and Perplexity both directed the user to premium tax refund companies, which are known for charging high fees and adding extra charges.
Andrew Laughlin of Which? advised consumers to "make sure to define your questions" when using AI to ensure they receive accurate and reliable information.
