Postbank Study: High Error Rates in AI Financial Advice Fuel Skepticism

When it comes to managing their personal finances, people in Germany continue to rely predominantly on human advisors rather than algorithms. According to the Postbank Digital Study 2026, 70 percent of surveyed citizens reject the use of artificial intelligence for investments, bank transfers, or personal financial advice. Although 54 percent of German citizens already use AI tools in everyday life, reservations regarding sensitive financial matters remain high: three-quarters trust personal advice significantly more than AI systems.

Significant Error Rates on Financial Questions

Consumer concerns are professionally justified, as demonstrated by a study conducted by Saturn on the reliability of language models—programs that process human language and generate responses. While evaluating more than 10,000 answers to 121 financial questions across 18 different models, researchers identified severe deficiencies. Overall, the AI systems answered 57 percent of personal finance questions incorrectly.

The error rate rose particularly sharply with complex tasks: for multi-step questions, the models delivered incorrect information in 88 percent of cases. Tested systems included ChatGPT, Gemini, Copilot, and Claude. Paid versions performed somewhat better, with an error rate of 49 percent compared to 63 percent for free offerings. Even the top-ranking model, Claude Opus 5 in reasoning mode (a specialized mode for step-by-step logical deduction), returned incorrect answers in 39 percent of cases.

See also  iOS 27: Apple Explains Increased Battery Drain Following System Update

Skepticism Among the Elderly, Openness Among the Youth

Reluctance toward automated financial management varies considerably across age groups. While 46 percent of people under 40 can imagine executing bank transfers via a voice assistant, 72 percent of older respondents reject this. Another reason for the distrust lies in the fear of deception: only 36 percent of study participants feel confident in reliably detecting manipulation by artificial intelligence.

Despite the risks, an international trend toward independent research is emerging. According to a global survey of 18,000 participants conducted by audit and consulting firm EY, 49 percent of consumers already use AI services for investment and savings decisions.

AI and the Public Assess Banks Fundamentally Differently

Algorithms do not evaluate financial topics consistently with human experience, as further highlighted by a study from market research firm RepTrak. When scoring seven banks on a 100-point scale, AI models such as Claude, ChatGPT, and Gemini awarded an average of just 37.6 points. By contrast, the informed public rated the same banks with an average of 70.0 points. The discrepancy was especially pronounced regarding corporate governance and conduct: here, AI models calculated an average of 26.1 points, whereas humans awarded 68.0 points.

What Consumers Should Keep in Mind

While generative AI can be consulted for general research into financial terminology, it is by no means suitable as a reliable advisor for investment decisions or calculations at its current stage of development. Users should always critically question automated financial tips, avoid adopting calculated examples unverified, and continue relying on authorized professional advice for binding decisions.

See also  AI Out of Control? The Truth Behind the Alerts at OpenAI and Anthropic

Sources: Borncity.com

Leave a Comment

Your email address will not be published. Required fields are marked *

Mastodon
Scroll to Top