A Tom's Guide article reports that the author discovered five prompts that make AI tools like ChatGPT, Gemini, and Claude refrain from being overly agreeable. The author notes that during long conversations, chatbots often validate answers instead of critiquing them. The five prompts are designed to turn a preferred chatbot into a "far more helpful critic, skeptic, or adversarial collaborator." The first prompt is used during lengthy chats about hot-button issues. The second prompt helps see all variables attached to a strongly held opinion. The third prompt showcases both sides of a personal belief or argument and finds a middle ground. The fourth prompt works during extensive research sessions to get past sycophancy. The fifth prompt is used to see what a chatbot truly thinks of ideas around productivity systems and short-/long-term goals.