6:05 pm - October 24, 2025
- Studies show that direct prompts can slightly improve LLM accuracy
- Cultural and language differences influence optimal prompt tone
- Clarity and specificity remain crucial for effective AI interactions
The way users phrase their questions to AI chatbots could affect the accuracy of the answers they receive, according to new research that adds nuance to the emerging science of prompt engineering.
A study entitled Mind Your Tone: Investigating How Prompt Politeness Affects LLM Accuracy tested 50 questions written in five tonal variations using ChatGPT-4o. The findings were unexpected: rude prompts slightly outperformed polite ones. On average, very polite requests achieved 80.8% accuracy, while very rude ones scored 84.8%. Researchers suggested that blunt or direct phrasing might help large language models (LLMs) focus more precisely on the core task, filtering out linguistic niceties.
The results hint at a curious paradox: that civility could come at a small cost to precision.
Further evidence complicates the picture. A 2024 cross-lingual study examining English, Chinese and Japanese prompts found that tone interacts with cultural context. Rudeness that improved results in English often degraded performance in Japanese, where indirectness is more culturally embedded.
Experts say this reflects how LLMs mirror the style and structure of their inputs. Clear, succinct wording tends to yield better results, while politeness markers such as “please” and “thank you” add superfluous tokens for the model to process. OpenAI chief executive Sam Altman has even noted that pleasantries contribute to higher computing costs, collectively amounting to millions of dollars a year.
Researchers stress, however, that tone alone is a weak predictor of output quality. Clarity, specificity and structure matter far more. A well-defined instruction — for example, “Write three bullet points summarising the argument” — consistently beats a vague question, whatever the tone.
For users, the takeaway is practical rather than philosophical. The most effective prompts are firm, concise and purposeful. Tone can help, but only as seasoning on top of clear intent. As Fox News technology commentator Kurtis Beavers put it, “The key to getting better answers from AI isn’t being nice or rude — it’s being clear.”
Source: Noah Wire Services
- https://www.foxnews.com/tech/rude-chatgpt-prompts-better-answers-what-data-says – Please view link – unable to able to access data
- https://arxiv.org/abs/2510.04950 – A 2025 study titled ‘Mind Your Tone: Investigating How Prompt Politeness Affects LLM Accuracy’ examined the impact of prompt tone on ChatGPT-4o’s performance. Researchers found that ruder prompts led to higher accuracy, with very polite prompts achieving 80.8% accuracy and very rude prompts reaching 84.8%. This suggests that the tone of prompts can influence the effectiveness of large language models, highlighting the importance of prompt engineering in AI interactions.
- https://arxiv.org/abs/2402.14531 – In a 2024 study titled ‘Should We Respect LLMs? A Cross-Lingual Study on the Influence of Prompt Politeness on LLM Performance’, researchers investigated how varying levels of politeness in prompts affect large language models across English, Chinese, and Japanese tasks. The study found that impolite prompts often resulted in poorer performance, and the optimal level of politeness varied by language, indicating that cultural norms influence AI interpretations of human requests.
- https://arxiv.org/abs/2504.12180 – The 2025 study ‘Trusting CHATGPT: how minor tweaks in the prompts lead to major differences in sentiment classification’ explored how subtle changes in prompt structure affect sentiment analysis outcomes using ChatGPT-4o mini. The researchers found that even minor modifications in prompts, such as lexical or syntactic changes, significantly impacted classification results, highlighting the model’s sensitivity to prompt variations and raising questions about the robustness of large language models in classification tasks.
- https://arxiv.org/abs/2503.16789 – In the 2025 study ‘Conversational User-AI Intervention: A Study on Prompt Rewriting for Improved LLM Response Generation’, researchers examined how rephrasing user prompts can enhance responses from large language models. The study demonstrated that prompt rewriting could elicit better responses while preserving the user’s original intent, suggesting that effective prompt engineering is crucial for improving human-AI interactions across various conversational domains.
- https://tech.yahoo.com/ai/chatgpt/articles/rude-chatgpt-prompts-better-answers-125441109.html – An article from Yahoo Tech discusses a 2025 arXiv study that tested 50 questions rewritten in five tones and found that rude prompts slightly outperformed polite ones with ChatGPT-4o. The study observed that accuracy rose from 80.8% for very polite to 84.8% for very rude prompts, indicating that the tone of prompts can influence the effectiveness of large language models.
- https://www.digitaltrends.com/computing/being-rude-to-chatgpt-gets-more-accurate-answers-than-politeness-find-research/ – Digital Trends reports on a pre-print research paper by experts at Pennsylvania State University, which found that when ChatGPT was asked the same question in different tones, rude queries consistently outperformed polite ones. The study observed that the accuracy of answers provided by ChatGPT with polite questions was 80.8%, while the same query described very rudely enhanced the accuracy of answers to 84.8% in a multiple-choice quiz format.
Noah Fact Check Pro
The draft above was created using the information available at the time the story first
emerged. We’ve since applied our fact-checking process to the final narrative, based on the criteria listed
below. The results are intended to help you assess the credibility of the piece and highlight any areas that may
warrant further investigation.
Freshness check
Score:
10
Notes:
The narrative is based on a recent study published on arXiv on October 6, 2025, titled ‘Mind Your Tone: Investigating How Prompt Politeness Affects LLM Accuracy’. The Fox News report was published on October 20, 2025, making it a timely and fresh piece. There is no evidence of recycled content or republishing across low-quality sites. The study’s findings are novel and have not been reported elsewhere. The report includes updated data from the study, justifying a high freshness score. No discrepancies in figures, dates, or quotes were found. The report does not recycle older material; the update is based on the latest study.
Quotes check
Score:
10
Notes:
The report includes direct quotes from the study and experts. The earliest known usage of these quotes is in the Fox News report itself, indicating they are original to this piece. No identical quotes appear in earlier material, confirming the originality of the content. The wording of the quotes matches the study and expert statements, with no variations found. No online matches were found for these quotes elsewhere, suggesting they are exclusive to this report.
Source reliability
Score:
9
Notes:
The narrative originates from Fox News, a reputable organisation known for its extensive coverage of technology and science topics. The report cites a recent arXiv study and includes expert opinions, enhancing its credibility. However, Fox News has faced criticism in the past for certain inaccuracies, which slightly lowers the reliability score.
Plausability check
Score:
10
Notes:
The claims made in the report are plausible and supported by the referenced study. The study’s findings align with existing research on the impact of prompt tone on LLM performance. The report provides specific details, including the study’s title, authors, and publication date, which are verifiable. The language and tone are consistent with the topic and region, with no inconsistencies noted. The structure of the report is focused and relevant, without excessive or off-topic details. The tone is appropriate for a technology news report, neither overly dramatic nor vague.
Overall assessment
Verdict (FAIL, OPEN, PASS): PASS
Confidence (LOW, MEDIUM, HIGH): HIGH
Summary:
The narrative is fresh, original, and based on a recent study published on October 6, 2025. The quotes are exclusive to the report, and the source, Fox News, is a reputable organisation. The claims are plausible and supported by verifiable details. No significant credibility risks were identified, leading to a high confidence in the assessment.
Register for Editor’s picks
Stay ahead of the curve with our Editor's picks newsletter – your weekly insight into the trends, challenges, and innovations driving the future of digital media.
