A new international study found that artificial intelligence interprets Polish more effectively than any other tested language. English, often considered the global standard, landed only in sixth place. Researchers from the University of Maryland and Microsoft compared 26 languages to measure how well AI systems understand and respond to prompts.
The results challenged expectations. “English did not perform at the top,” the research team noted. “It ranked sixth, while Polish delivered the highest accuracy across long-text evaluations.”
Machines Excel Where Humans Struggle
The researchers examined responses from several major AI systems, including OpenAI, Google Gemini, Qwen, Llama, and DeepSeek. Each system received identical tasks translated into 26 different languages.
Polish reached an impressive 88% success rate, outperforming every other language tested. The Polish Patent Office reacted to the findings online, writing, “AI follows Polish commands with remarkable precision. People find Polish hard to master, but AI handles it with ease.”
Despite limited Polish-language data available for training, AI systems demonstrated advanced comprehension. In contrast, Chinese performed unexpectedly poorly, ranking near the bottom of the list.
Global Rankings of AI Language Accuracy
The study produced a clear ranking of how well AI systems process different languages:
- 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 understanding does not depend on how widely a language is spoken or how complex it is. Paradoxically, one of the hardest languages for humans turned out to be the easiest for machines.
