Researchers discovered that artificial intelligence understands Polish better than any other language, placing English only sixth. A joint study by the University of Maryland (UMD) and Microsoft analyzed 26 languages to determine which ones AI models process most effectively.
“Our results surprised us,” the report’s authors stated. “English did not dominate every model. It ranked sixth out of 26 languages when handling long texts, while Polish ranked first.”
The research team tested major AI systems, including OpenAI, Google Gemini, Qwen, Llama, and DeepSeek, by giving them identical tasks in 26 languages. Polish achieved an impressive average accuracy of 88% in completing those tasks.
AI Handles Polish Commands with Unusual Precision
The analysis showed that Polish produced the most accurate commands for AI models. The Polish Patent Office highlighted the finding on Facebook, writing, “Humans may struggle with Polish, but AI does not.”
Despite limited Polish-language training data compared to English or Chinese, AI systems still demonstrated remarkable proficiency. This suggests that AI models can adapt well to linguistic complexity, even when data availability is lower.
In sharp contrast, Chinese performed poorly, ranking fourth from the bottom among all tested languages. Researchers described the result as unexpected given the vast quantity of Chinese-language data used to train most models.
English Trails Behind as Europe Dominates Rankings
European languages dominated the top 10 list for conversational AI efficiency. The ranking placed Polish at the top, followed by French at 87%, Italian at 86%, and Spanish at 85%. Russian came next at 84%, while English followed at 83.9%. Ukrainian reached 83.5%, Portuguese 82%, German 81%, and Dutch 80%.
The study challenges assumptions about English’s supremacy in AI communication. It also highlights that linguistic complexity and structure, rather than data volume alone, may play a crucial role in how effectively AI interprets human language.
