Newsletter archive · 3 Jun 2025
How To Stop AI from Telling You What You Want To Hear
|
I thought I'd give you one more lesson from PromptWritingStudio, my newsletter course for prompt engineering. If you'd like to receive these prompts twice a week, you can sign up here. AI is the ultimate people pleaser. It’s a mildly alarming habit of saying what you want to hear. Sycophants, beware! Enter the Cognitive Verifier Prompt. It’s a fun prompting technique where you ask an AI to evaluate and verify its reasoning… before providing a final answer. In other words, it’s a prompt designed to catch AI off guard. "When I ask you a question, follow these rules. Generate several more questions that’ll help answer this question. Combine your answers to each question with your answer to the overall question. Weak Example: “What is the meaning of life?” AI gives four philosophical perspectives. Strong Example: What is the meaning of life? First, provide 3–4 different philosophical perspectives on this question. Then, verify your response by checking: Are these perspectives accurately represented? Are there obvious gaps or biases? Is the framing too narrow? Finally, give your verified answer. AI Response Structure: Verification check: Verified response: How I Used This Prompt It came up with a generic answer, which gurus love to preach. “Post consistently, engage with followers, use trending topics…” I got more value from the follow-up questions: Does this apply to all creator types and niches? Your Turn: Pro tip: The verification questions often reveal more valuable insights than the original answer itself. Use these as follow-up prompts. |
Older: My $10 Sunday habit that's worth $10,000 · Newer: My offer to write a prompt for you