Two images One is real One is an AI influencer with 121K+ followers Comment 1 or 2 below 👇 I’ll reveal the answer later this evening and break down what gives it away This is how realistic AI visuals have become

📝 TL;DR A new research paper finds that rude prompts can make ChatGPT significantly more accurate on test questions. The twist, it is not about being a jerk, it is about how blunt, low fluff language makes your instructions clearer to the model. 🧠 Overview Researchers from Penn State tested how different tones, from very polite to very rude, affect ChatGPT 4o’s accuracy on multiple choice questions in maths, science, and history. Surprisingly, the ruder prompts consistently scored higher than the polite ones. This challenges the idea that you should always be extra polite to get the best answers from AI and instead points to clarity and directness as the real performance drivers. 📜 The Announcement The paper, titled “Mind Your Tone: Investigating How Prompt Politeness Affects LLM Accuracy,” rewrote 50 base questions into five tone variants, Very Polite, Polite, Neutral, Rude, and Very Rude, for a total of 250 prompts. The team ran all of these through ChatGPT 4o and compared how often the model chose the correct answer. Very polite prompts scored about 80.8 percent accuracy, while very rude prompts scored about 84.8 percent, a roughly four point jump that was statistically significant. The authors note that this result flips what earlier studies found, where rude prompts often hurt performance, which suggests that newer models may react differently to tone. ⚙️ How It Works • Five tone versions per question - Each of the 50 questions was rewritten in Very Polite, Polite, Neutral, Rude, and Very Rude styles so the content stayed the same but the tone changed. • Same model, same questions, different tone - Only the tone wrapper changed, all prompts were sent to ChatGPT 4o, so differences in accuracy could be linked to tone rather than content. • Rude prompts remove “politeness padding” - The ruder prompts tended to be shorter, more direct, and less hedged, which means less extra text for the model to parse. • Polite prompts add linguistic noise - Very polite wording often included extra phrases like “would you kindly” or “if it is not too much trouble,” which may dilute the core instruction.

Hello everyone! I am Morris, an entrepreneur from South Sudan. I have been using ChatGPT for several months, but I am still seeking a clearer way to navigate it. I am open to guidance from experts who can help me learn more about what’s possible inside this space. AI feels like a new environment am learning to align with it in to my life and work. "It is true we can gauge the river only after we jump in".