How Well Do LLMs Know Double-Entry Accounting? (www.moderntreasury.com)

A recent study explored the capabilities of large language models (LLMs) in performing double-entry accounting tasks, particularly in generating charts of accounts and ledger transactions. The research tested both open-source and frontier models using a systematic command-line tool that assessed their accuracy, balance, and adherence to accounting principles. The findings revealed that while LLMs can assist both novices and experts in chart creation, their effectiveness varies significantly among models. Notably, OpenAI’s GPT model emerged as the top performer, demonstrating strong understanding in constructing accurate accounts and generating balanced transactions. This exploration is significant for the AI/ML community as it highlights the potential of LLMs to bridge knowledge gaps in specialized fields like accounting. The study not only provides a framework for evaluating model performance but also emphasizes the importance of structured outputs in achieving accounting correctness. While some models struggled with basic concepts—like balance correctness and normality of accounts—frontier models like Gemini 3 and GPT-5.2 excelled across various scenarios, indicating that improvements are continuously being made in AI’s capacity to tackle complex business tasks effectively.