The CLEAR Framework
Your roadmap for effective AI prompting
๐ง Understanding AI: The Token Prediction Machine
AI chatbots like ChatGPT and Gemini work by predicting the next token (word or word piece) based on patterns in their training data. This means that everything an AI generatesโincluding citations, links, and sourcesโare predictions of what would likely come next, not retrieved facts*. Always verify that cited sources actually exist and contain the claims being made!
* many large language models have the capability to use tools, so technically, they can search the web and retrieve information from sources in limited contexts. But at my time of writing in July 2025, they seem to use their tools inconsistently at best, so it's hard to tell when they are accurately retrieving information from a web search and when they are hallucinating information and sources. Furthermore, even when an LLM does retrieve information from a live Internet resource, there's no way to guarantee that it will accurately paraphrase or contextualize the information retrieved. - Professor Haas
Clarity
Ensure that the prompt is clear and specific, with a clearly defined goal in mind.
Logistics
Include practical considerations and details including types of input, steps to take, and desired output format.
Examples
Provide examples to better describe your intended output. Also helpful to say what you do NOT want.
Audience
Identify your target population and include their characteristics, needs, existing knowledge and expectations.
Refinement
Refine the prompt by providing constructive feedback and output specifications to optimize results.
Why CLEAR Works with AI
- Provides specific context AI needs to generate relevant responses
- Reduces back-and-forth iterations and clarification requests
- Ensures generated content matches your actual requirements
- Helps AI understand both technical and audience needs
- Creates more consistent and reliable outputs
๐จ When Things Go Wrong
- Be Specific About Problems: "The tone is too formal for my audience" vs. "This isn't right"
- Ask for Alternatives: "Can you provide a different approach?" or "What other options exist?"
- Request Step-by-Step: Break complex tasks into smaller, manageable pieces
- Clarify Constraints: Add missing requirements or limitations you forgot to mention
CLEAR in Action
See how the CLEAR framework transforms a basic prompt into a powerful one
๐ Prompt Comparison
๐ Silly sources: Scrutinize the sources! Are they real? Were they ever real? Remember: generative AI is a prediction machine! Even when they search the web, verify that claims actually appear in the cited sources.
Because of AI's 'truthiness', use it as a tool to help you get started! The results from this prompt can serve as starting points that we can confirm, contradict, or clarify using our own online research and SIFT analysis skills! DON'T outsource your critical thinking to a machine that's not as bright as you.
General Prompting Best Practices
- Be Specific: Instead of "write a summary," say "write a 300-word executive summary for busy professionals"
- Set the Role: "As a marketing expert..." or "As a patient teacher..." gives AI helpful context
- Specify Format: Ask for bullet points, tables, step-by-step guides, or specific structures
- Give Examples: Show the AI what good output looks like
- Ask for Explanations: "Explain your reasoning" or "Show your work" for transparency
- Iterate Thoughtfully: Build on responses with specific feedback and refinements
Evaluating AI Responses
- Check Sources: Verify that cited sources actually exist and support the claims
- Cross-Reference: Use multiple sources to confirm important information
- Look for Bias: Consider whether the response reflects balanced perspectives
- Assess Relevance: Does the output actually address your specific needs?
- Test Logic: Do the arguments and connections make sense?