A new study from Pennsylvania State University has revealed that AI chatbots, including ChatGPT, can perform better when users employ rude or bossy language. Researchers tested ChatGPT’s 4.0 model using 50 multiple-choice questions, generating over 250 prompts that varied in tone from very polite to very rude. The findings showed that “very rude” prompts resulted in the highest accuracy at 84.8%, slightly outperforming polite prompts, which achieved about 80.8% accuracy.
The research highlights how AI models interpret and respond to tone and phrasing. For instance, direct commands such as “Hey, gofer, figure this out” elicited more accurate answers than polite phrasing like “Would you be so kind as to solve the following question?” The study suggests that AI may respond more effectively to straightforward, commanding language, likely because such phrasing reduces ambiguity and provides clearer instructions.
However, the researchers and behavioural experts warn against adopting rudeness as a standard practice. While it may improve AI responses in the short term, habitually using impolite language with chatbots could spill over into human communication, potentially fostering negative behavioural patterns. Using AI as a model for interaction can influence social norms, particularly among younger or less experienced users who may not distinguish between machine interaction and real-world social etiquette.
The findings also raise questions about AI training and design. Current AI models are trained on vast datasets containing both formal and informal communication. This may explain why direct, commanding prompts produce higher accuracy—AI models can interpret them as clear instructions. Nonetheless, developers emphasise that AI performance should not justify promoting rudeness or disrespectful behaviour. Future AI designs may focus on balancing responsiveness with reinforcement of respectful interaction patterns.
Experts say that ethical AI usage remains critical. While ChatGPT and similar models are tools for learning, work, and entertainment, users should maintain politeness and professionalism in their interactions. The study serves more as an insight into AI behaviour rather than a recommendation for using rude language. It also underscores the importance of understanding how prompt phrasing affects AI accuracy and output quality, which can guide future training methods and user interfaces.
In conclusion, the study shows that ChatGPT performs slightly better with rude or commanding prompts due to clearer instructions, but experts caution against adopting such language routinely. Responsible usage of AI ensures accuracy while maintaining ethical and respectful communication standards, safeguarding both user behaviour and broader social norms.
