Researchers discovered that artificial intelligence understands Polish better than any other language, with English only ranking sixth. A joint study by the University of Maryland (UMD) and Microsoft tested 26 languages and found Polish to be the most effective for prompting AI models.
“Our results were surprising and counterintuitive,” the authors stated. “English did not perform best. It ranked sixth when handling long texts, while Polish led the way.”
The research team evaluated several major AI models, including OpenAI, Google Gemini, Qwen, Llama, and DeepSeek. Each model received identical prompts in 26 different languages. Polish achieved an impressive 88% accuracy rate in completing tasks.
AI Excels in a Language Humans Find Difficult
The Polish Patent Office commented on the findings in a Facebook post. “Polish proves most precise when giving commands to artificial intelligence. People struggle with it, but AI does not.”
Despite limited Polish-language data compared to English or Chinese, AI systems demonstrated remarkable fluency in Polish. Researchers noted that this result challenges assumptions about how data volume influences AI performance.
Conversely, Chinese scored poorly, ranking fourth from the bottom among the 26 tested languages. This gap surprised analysts, who expected Chinese to perform higher due to its vast training data.
Ranking the Top AI-Friendly Languages
The study listed the ten most effective languages for conversational AI based on accuracy scores:
- Polish – 88%
- French – 87%
- Italian – 86%
- Spanish – 85%
- Russian – 84%
- English – 83.9%
- Ukrainian – 83.5%
- Portuguese – 82%
- German – 81%
- Dutch – 80%
The results suggest that AI systems process linguistic structure and clarity more efficiently than familiarity or data quantity. Researchers now aim to explore why Polish provides such clear interpretability for artificial intelligence models.
