A new global study found that artificial intelligence responds most accurately to prompts written in Polish, with English ranking only sixth. The University of Maryland and Microsoft tested 26 languages across several major AI systems to determine which languages machines process most effectively.
Researchers expressed surprise at the findings, noting that English—often considered the default AI language—did not lead. “English finished sixth when we evaluated longer texts, while Polish produced the strongest performance overall,” the report stated.
The study involved AI models from OpenAI, Google Gemini, Qwen, Llama, and DeepSeek, each tested with identical instructions translated into different languages.
Machines Excel Where Humans Struggle
Polish delivered an average task accuracy of 88%, placing it at the top of the list. The Polish Patent Office noted, “Polish turns out to be the clearest for AI commands. People may find it hard to learn, but AI seems to master it easily.”
Researchers found this result remarkable given the limited amount of Polish data used in AI training compared with English or Chinese. Despite its smaller dataset, AI handled Polish prompts with exceptional precision.
Chinese, on the other hand, underperformed, ranking fourth from the bottom among the 26 languages tested.
The Top Ten Languages for AI Precision
The experiment ranked the most effective languages for AI comprehension and response:
- Polish — 88%
- French — 87%
- Italian — 86%
- Spanish — 85%
- Russian — 84%
- English — 83.9%
- Ukrainian — 83.5%
- Portuguese — 82%
- German — 81%
- Dutch — 80%
The researchers concluded that AI doesn’t always perform best in the languages humans find easiest. Paradoxically, the world’s most challenging languages may help machines think with the greatest clarity.
