The use of artificial intelligence today is transforming the way people work, communicate and receive information. Many are polite to chatbots to get the right answer for them, and many others “threaten” artificial intelligence to be accurate and precise based on the user’s needs. “A lot of people think there’s a magic set of words you can use that will make big language models solve a problem, but it’s not about the choice of words, it’s about how you essentially express what you’re trying to do,” said Jules White, a professor of computer science who studies generative AI at Vanderbilt University in the US.
This means that everything you say, from your choice of words to an extra comma, will affect how the AI responds. The problem is that it’s incredibly difficult to predict.
According to a BBC article, a 2024 study showed that large language models gave better and more accurate answers when asked politely rather than simply giving orders. Even more surprising was that there were cultural differences. Compared to Chinese and English, Japanese-speaking chatbots actually performed slightly worse if you were a little more polite. Although this study was conducted in 2024, things have changed since then. Today, Rick Battle, an applied machine learning engineer at Broadcom, and others say that the newer AI models you find in every major product like ChatGPT, Gemini, or Claude are better at capturing the most important parts of your request.
Companies design artificial intelligence like ChatGPT or Google’s Gemini to behave like humans, so it makes sense that they can sometimes seem like they have a sense of humor or personality.
AI tools are imitations, not living beings. They simply simulate human behavior. If you want better answers, stop treating AI like a person and start treating it like a tool. “The first thing I tell people is not to ask for one answer, but three or five,” says White. Give the AI a sample whenever possible. White says it’s much more effective to say, “Here are 10 emails I’ve sent in the past, using my writing style.” It’s also worth noting that when asking for information or asking questions with a specific answer, experts say role-playing can make AI models less accurate.
As for politeness, a 2025 survey by publisher Future found that 70% of people are polite to artificial intelligence when they use it. Most said they are polite because it is the right thing to do, although 12% said they do it to protect themselves in the event of a robot uprising.
“The most important thing for me is that saying ‘please’ and ‘thank you’ can make you feel more comfortable interacting with artificial intelligence. It’s not helping the model’s performance, but if it’s helping you use the model more because you feel more comfortable, then it’s helpful,” said Sander Schulhoff, an entrepreneur and researcher who helped popularize the idea of agile engineering.