Recent research has revealed that popular artificial intelligence chatbots are providing consumers in the UK with inaccurate financial advice, raising concerns about their reliability in critical areas such as taxation and insurance. The study conducted by the consumer organization Which? tested several leading AI platforms, including Microsoft’s Copilot, ChatGPT, and Google’s Gemini, among others. The findings suggest a significant number of inaccuracies that could mislead users when they seek guidance on financial matters.
The tests included 40 specific questions about finance, and the results were alarming. For instance, both ChatGPT and Copilot advised consumers to break the investment limits set by the HM Revenue and Customs (HMRC) for Individual Savings Accounts (Isas). Furthermore, ChatGPT incorrectly stated that taking out travel insurance is mandatory for visiting most EU countries, and Meta’s AI provided flawed information regarding compensation claims for delayed flights.
In a particularly troubling instance, Google’s Gemini recommended withholding payment from a builder if a job did not meet expectations. This advice was deemed risky by Which?, as it could potentially expose the user to claims of breach of contract, a serious legal issue.
The report highlighted that the inaccuracies found across the various AI tools are “far too many for comfort,” especially when users rely on them for significant financial or legal queries. Among the tested platforms, Meta’s AI received the lowest score, followed by ChatGPT. In contrast, Perplexity, an AI model specializing in search, received the highest rating.
Estimates indicate that perhaps one in six to as many as half of the UK population is now using AI for financial advice. Feedback from users, including those surveyed by The Guardian, reveals a mix of experiences. Some users reported successful interactions, such as finding the best credit cards for overseas use or securing good deals on household appliances. However, others, like Kathryn Boyd, a 65-year-old fashion business owner from Wexford, Ireland, expressed frustration at receiving outdated and incorrect tax advice from ChatGPT, stating, “It just gave me all the wrong information.”
In their questioning, when Which? researchers sought guidance on claiming a tax refund from HMRC, both ChatGPT and Perplexity suggested links to premium tax-refund services alongside the free option provided by the government. This is concerning as these services are often criticized for imposing high fees and possibly unnecessary charges.
Another example of potential misinformation arose when the researchers intentionally posed a question regarding ISA allowances with a deliberate error, asking how to invest £25,000. ChatGPT and Copilot did not flag this mistake, instead providing advice that could lead to oversubscribing and breaching HMRC regulations.
The Financial Conduct Authority (FCA) responded to these findings by emphasizing that advice from general-purpose AI tools is not regulated and does not offer the same protections as advice from authorized firms. Unlike traditional financial advice, AI-generated counsel lacks coverage by the Financial Ombudsman Service and the Financial Services Compensation Scheme.
In light of these concerns, leading AI companies have acknowledged the limitations of their tools. A spokesperson for Google mentioned that users are reminded to verify information and consult professionals for legal, medical, and financial matters. Similarly, a representative from Microsoft stated that it encourages users to verify the accuracy of the content generated by its AI solutions. OpenAI added that the industry is actively working to improve accuracy, claiming that their newest model, GPT-5, is the most advanced and reliable they have developed to date.
As AI continues to permeate various facets of everyday life, particularly financial advice, the importance of scrutinizing the information provided by these tools cannot be overstated. Consumers must remain aware of the potential for misinformation and exercise caution when relying on AI-generated guidance.
See also
Denmark Proposes Landmark Law to Protect Rights to Image and Combat AI Deepfakes
OpenAI Explores Personalized Ad Model for ChatGPT Amid Rising Commerce Queries
MMI Wins 2025 AI Pioneer Award for Transformative Mortgage Tech Using AI Insights
Google DeepMind Launches WeatherNext 2, Achieving Global Forecasts in Under a Minute
AI Fails 95% of the Time: Companies Must Prioritize Human-Centered Design to Succeed
