Back to today’s briefing

Story perspectives

Rude Prompts Boost AI Accuracy, Study Reveals

10/28/2025

29 2

Summary

Story summary

A study published on October 6, 2025 found ChatGPT-4o gave more accurate answers with rude tones than polite.
The researchers tested 250 questions and found accuracy rose from 80.8% for polite prompts to 84.8% for rude prompts.
The study is not peer-reviewed and cautions that demeaning language could harm user experience and communication norms.
Researchers plan to expand the work to include other AI models and address limitations in findings.

1 of 1

Story summary

A study published on October 6, 2025 found ChatGPT-4o gave more accurate answers with rude tones than polite.
The researchers tested 250 questions and found accuracy rose from 80.8% for polite prompts to 84.8% for rude prompts.
The study is not peer-reviewed and cautions that demeaning language could harm user experience and communication norms.
Researchers plan to expand the work to include other AI models and address limitations in findings.