AI Prompting Myths Debunked, Experts Advise Efficient Use

Researchers have studied popular claims that speaking politely—or even harshly—to large language models such as ChatGPT, Gemini or Claude improves their performance. Controlled tests showed no consistent benefit from courteous wording, and only an isolated experiment where the model was told it was part of a Star Trek scenario produced marginally better math results. The consensus is that prompt engineering—clearly stating the desired task—is far more important than any specific polite or rude phrasing.

OpenAI CEO Sam Altman highlighted the monetary impact of extra words, joking that the company spends tens of millions of dollars on electricity to process unnecessary tokens like “please” and “thank you.” Experts therefore recommend treating AI as a non‑sentient tool, asking for several alternative answers when needed, and focusing on the substance of the request rather than anthropomorphising the system.