No. 36 Please & thank you.
Where it comes from
In 2024, Yin et al. tested polite & rude prompts in English, Chinese & Japanese. Rude ones often did worse, but extra politeness didn’t reliably help.
In 2025, Dobariya & Kumar tested ChatGPT-4o on 50 questions. Very rude prompts scored 84.8%, very polite ones 80.8%.
Small studies, different models. The honest answer: it depends & manners are free.
- Yin, Wang, Horio, Kawahara & Sekine, “Should We Respect LLMs? A Cross-Lingual Study on the Influence of Prompt Politeness on LLM Performance” (2024, arXiv:2402.14531)
- Dobariya & Kumar, “Mind Your Tone: Investigating How Prompt Politeness Affects LLM Accuracy” (2025, arXiv:2510.04950)
