A new study found that artificial intelligence interprets Polish far more effectively than any other language tested. English, often assumed to be the default for AI, came in only sixth. Researchers from the University of Maryland and Microsoft evaluated how well top AI models processed the same inputs across 26 languages.
The results defied expectations. “English failed to perform best,” the authors explained. “It ranked sixth, while Polish showed the highest accuracy.” Major systems like OpenAI, Google Gemini, Qwen, Llama, and DeepSeek all participated in the comparison.
Machines Excel Where Humans Struggle
Polish achieved an impressive 88% average accuracy, marking it the most precise language for giving AI instructions. The Polish Patent Office celebrated the outcome on Facebook, noting, “Polish challenges people but not machines.”
Despite limited Polish data available for AI training, the models handled it better than languages with vastly larger datasets. Chinese, for instance, performed poorly, finishing near the bottom of the rankings. The discovery suggests that language structure, not data volume, may influence AI comprehension.
Languages Ranked by AI Accuracy
The research ranked the ten clearest languages for AI interaction as follows:
- 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 systems follow logic over linguistic difficulty. In this case, the machines mastered one of humanity’s toughest languages with ease, overturning long-held assumptions about language learning—human or artificial.
