AI Models and Financial Data Accuracy
A study conducted by Reputation Age and Bubbling, released on October 2, 2026, has unveiled a critical issue in the application of artificial intelligence (AI) for financial analysis. The study focused on the accuracy of AI models, specifically ChatGPT and Gemini, in interpreting the semi-annual financial figures of 39 companies listed on the CAC 40, a major French stock market index.
Key Findings
- Error Rate: The study found that 40% of the financial figures provided by these AI models were incorrect.
- Accuracy Variability: The accuracy of AI-generated data varied significantly among companies, with Capgemini achieving an 81% accuracy rate, while Crédit Agricole lagged at 32%.
- Source of Errors: A notable portion of errors stemmed from the AI's inability to extract the correct indicator from the right document, despite having access to accurate source material.
Implications for Financial Analysis
The findings underscore a critical issue in the reliability of AI for financial analysis, particularly in the context of the CAC 40. The potential for misinterpretation of financial data poses significant risks for decision-making processes that rely on AI-generated insights.
Actors and Market Impact
- Gemini: As a player in the crypto sector, Gemini's involvement in this study highlights the broader implications of AI inaccuracies beyond traditional finance.
- Bubbling and Reputation Age: These entities co-authored the study, bringing attention to the need for improved AI accuracy in financial contexts.