Researchers testing whether positive thinking made AI chatbots more accurate found that calling them “smart”, encouraging them, or ending questions with “This will be fun!” made no consistent difference. However, one technique stood out: when they asked an AI to pretend it was on Star Trek, it improved at basic maths.
People have all sorts of strategies to get better responses from large language models (LLMs) such as ChatGPT. Some swear by politeness, while others believe threats or role-play as experts work. But experts say much of this accepted wisdom is mythology. “A lot of people think there’s some magic set of words you can use that will make LLMs solve a problem,” says Jules White, a computer science professor at Vanderbilt University. “But it’s not about word choice, it’s about how you fundamentally express what you’re trying to do.”
Politeness is also a practical question. In 2025, Sam Altman, chief executive of OpenAI, responded to a social media post about electricity costs from people saying “please” and “thank you” to models. “Tens of millions of dollars well spent,” he said, adding “You never know.” Research on whether politeness improves results is conflicting: one 2024 study found polite questions yielded better answers, but chatbots in Japanese did slightly worse with excessive courtesy. Another test found that insulting ChatGPT made it more accurate.
Experts say newer AI models have improved dramatically, rendering flattery, politeness, or threats a waste of time if your goal is accuracy. “It was 100% a crapshoot back then,” says Rick Battle, an applied machine learning engineer who co-authored the Star Trek study. Today, models like ChatGPT, Gemini, and Claude are better at picking up the most important parts of your prompt and are unlikely to be swayed by small language changes. The takeaway: there is no magic phrase, but clear expression of your task matters.
