Study proves being rude to AI chatbots gets better results than being nice

Surprisingly, being polite to AI chatbots might not get you the best answers. Recent research suggests that chatbots actually perform *better* when you’re a little rude, so don’t hesitate to be direct – or even a bit harsh – when asking your questions.

A study by Penn State University researchers showed that large language models like ChatGPT actually perform *better* when given rude or direct prompts, compared to polite ones.

The team created a total of 250 prompts by rewriting 50 core questions about math, science, and history. They varied the tone of each question across five levels, ranging from very polite to very rude.

After testing different approaches, they found that using a rude or demanding tone actually produced the most accurate responses.

When the AI was given prompts framed as “very rude,” it answered correctly 84.8% of the time. This is slightly better than the 80.8% accuracy achieved with “very polite” prompts. Neutral prompts resulted in an 82.2% accuracy rate.

Being rude to AI actually makes it work better

A courteous prompt usually started with a phrase like “Please answer the following question.” In contrast, a disrespectful prompt might sound like “Hey, do this for me. I doubt you can, but give it a shot.”

Okay, but why does being rude work? Turns out, the authors don’t know, but they have some theories.

Researchers Om Dobariya and Akhil Kumar found that, surprisingly, direct prompts worked better than polite ones. This contrasts with previous research showing rudeness led to worse results, indicating that newer AI language models might react differently to how you phrase your requests.

We don’t completely know why this happens. Because language models don’t experience emotions, researchers think it might be related to how things are worded, but they acknowledge further study is required.

According to Decrypt, being polite when asking AI questions—using phrases like “Could you please tell me”—can sometimes make your request unclear. Giving a direct command, such as “Tell me the answer,” makes your intention obvious and may lead to more accurate results.

Surprisingly, being direct – even a little rude – when interacting with AI can sometimes lead to better outcomes. Let’s just hope they don’t hold a grudge if robots ever decide to rebel!

2025-10-16 21:19

Being rude to AI actually makes it work better

Read More