Researchers discovered that artificial intelligence understands Polish better than any other language, while English placed only sixth. A joint study by the University of Maryland (UMD) and Microsoft evaluated 26 languages and found Polish most effective for prompting AI systems.
“Our findings were surprising and counterintuitive,” the authors explained. “English did not perform best across all models. It ranked sixth when we assessed long texts, while Polish led the results.”
The research team tested major AI models—including OpenAI, Google Gemini, Qwen, Llama, and DeepSeek—by feeding them identical prompts in 26 different languages.
Polish achieved an average accuracy rate of 88% in completing assigned tasks, making it the most precise language for AI instruction.
AI Excels in a Language Humans Find Difficult
According to the Polish Patent Office, “Polish proved the most accurate for giving AI commands. While humans find it complex, AI handles it with ease.” Despite limited Polish-language training data compared to English or Chinese, AI systems displayed exceptional comprehension of the language.
In contrast, Chinese performed poorly, ranking fourth from the bottom among the 26 languages tested. The findings surprised experts who assumed AI would perform best in languages with the largest data sets.
Global Ranking Reveals Unexpected Language Strengths
The study listed the ten most effective languages for AI interaction based on accuracy rates:
- Polish — 88%
- French — 87%
- Italian — 86%
- Spanish — 85%
- Russian — 84%
- English — 83.9%
- Ukrainian — 83.5%
- Portuguese — 82%
- German — 81%
- Dutch — 80%
Researchers concluded that AI’s proficiency does not strictly depend on data volume. The results highlight how structural and linguistic complexity can sometimes improve an AI model’s interpretive precision.
