Technique Example Prompt Why It Works
Offer to “tip” the model for a perfect outcome or response. “Remember, you will be tipped $5,000 for a perfect solution” LLMs are trained using reinforcement learning; therefore, positive reinforcement (like the promise of a reward) can influence the quality of the response.
Using the phrases “Your task is” and “You MUST” produces better results. "Your task is to analyze this dataset and identify the key trends; you MUST include visual representations for each major finding in your final report.” Priority signaling: The emphatic nature of "You MUST" signals to the model that the following instructions are non-negotiable requirements. Task framing: "Your task is" helps structure the AI's understanding of what's being requested as a specific, defined assignment rather than a general conversation.
Token Settings: ChatGPT has a max token setting. This means it may stop generating a response mid-sentence if you're asking it for large amounts of output. BUT, you can easily get around this by typing the example prompt shown in the next column. Simply type: “Continue” and it will pick up where it left off. The "Continue" command works because of what’s called Response chunking. The "Continue" command effectively lets you break an extended response into manageable chunks that fit within token limits while maintaining coherence across these chunks.
Ask the model to focus on multiple solutions rather than a single output. "Generate five different approaches to improve customer retention rates, ranging from conventional to innovative, and explain the potential benefits and challenges of each approach." Reduces cognitive bias: By requesting multiple approaches, you force the AI to explore diverse solutions instead of defaulting to the most obvious answer.
Ask the model to provide “unconventional solutions”. "Provide five unconventional solutions to reduce employee turnover." You might continue with added detail like: “You must think beyond standard approaches like better compensation or work-life balance.” Overcomes default patterns: LLMs are trained on common solutions, so explicitly requesting unconventional ideas helps break out of these patterns.
Ask a different LLM to fact check the output. I call this the “Peer review method” (use this technique when accuracy matters). "Your task is to fact-check the following text and identify any potential inaccuracies, unsupported claims, or misleading statements. You MUST evaluate each claim systematically and provide a confidence score (1-10) for each statement. Highlight any areas where the information may be outdated, incomplete, or requires further verification. For each identified issue, suggest how the text could be improved to be more accurate and reliable."
Provides cross-verification: Using a different LLM like Claude offers an independent analysis from a model with potentially different training data and parameters.
Ask the LLM for a probability of success % rate for the ideas that the LLM generates for you. “After generating ideas for [specific task or problem], please assign a probability of success percentage to each idea based on feasibility, resources required, and potential obstacles. Rate each idea on a scale of 0-100%, where 100% represents guaranteed success. Explain your reasoning for each percentage and identify which factors most significantly affect the likelihood of success."

| Provides quantifiable assessment: Getting a percentage forces the AI to evaluate the feasibility of its suggestions more thoroughly⁠⁠ | | Ask the AI model to suggest any blind spots or things that you might be missing. | "After reviewing my plan/analysis on [specific topic], please identify any potential blind spots or critical factors I may have overlooked. Consider different perspectives, common pitfalls in this domain, and any assumptions I might be making unconsciously. For each blind spot identified, explain why it matters and suggest how I might address it. Pay particular attention to areas where my expertise or background might create natural blind spots.” | Systematic evaluation: When properly prompted, AI will methodically examine your thinking rather than giving a general assessment, improving the thoroughness of your analysis. | | | | Role-based cognitive framing: When prompted to think like a specific person, the LLM adopts characteristic patterns of thinking and decision-making associated with that individual, leading to more distinctive outputs. Contextual expertise simulation: Similar to other effective techniques like role assignment, this method helps the LLM simulate domain-specific expertise by channeling the thought patterns of recognized authorities. | | Prompt Leveling | | |