How to Write Better Prompts for AI

Artificial intelligence is powerful, but it's only as good as the instructions you give it. Think of prompting like writing a recipe: if you want the AI to "bake" something useful, you need to provide the right ingredients and clear instructions. Without them, you'll end up with guesswork instead of precision.

Why Prompting Matters

Prompts are recipes. The AI needs steps and details in a logical order to deliver the result you want.
AI makes mistakes. Sometimes it skips steps or invents information — these errors are called hallucinations. If accuracy matters, ask the AI to provide sources.
Context overload hurts. Just like scrolling through three pages of blog before you get to the recipe, too much irrelevant context confuses the AI. Include only what's necessary.

Imagine asking your AI to "Bake a chocolate cake with white frosting." Without ingredients or instructions, the AI has to improvise. That's exactly how it responds to vague prompts. The clearer your recipe, the better the cake.

The Three Essentials of a Good Prompt

Give an overview of the task. Example: "I need to create a document with these sections…"
Provide step-by-step instructions. Example: "First, add headings. Second, generate a table of contents. Third, summarize each section…"
Offer examples or watch-outs. Example: "When formatting the table, use column headings like: Name, Organization, Telephone."

AI DOs and DON'Ts

✅ DOs

Write clear, detailed instructions. Treat the AI like a new intern: eager but inexperienced.
Organize instructions logically with bullets or numbers.
Provide examples of preferred output. For instance: "Format all dates as mm/dd/yy. January 1, 2025 becomes 01/01/25."

❌ DON'Ts

Share personal information (emails, family names, sensitive data). Once entered, it can't be retrieved or deleted.
Blindly trust AI answers. Hallucinations happen, and they can sound convincing. Always verify.

How to Engineer an Effective Prompt

Step 1: Overview. "I want to create a report of CEOs for various companies."
Step 2: Logical steps. "For each company, return a CSV with headers: Name, Company, LinkedIn, Website."
Step 3: Test with one example. "Return the CEO details for Google."
Step 4 (Optional): Provide sample output, especially if you notice discrepancies or mistakes. Example: If one of the companies were Accelerated Motion Solutions, the proper output would be Michael Courter, Accelerated Motion Solutions, https://linkedin.com/in/michaeljcourter/, https://acceleratedmotion.com
Step 5 (Optional): Handle missing data. "If details are unavailable, mark the field N/A."

Adjusting for AI Limitations

AI tools have limits on how much text they can process or return. Push too far, and hallucinations increase. To work around this:

Break large tasks into smaller chunks.
Example: Instead of asking for 100 CEOs at once, split the list into sets of 25.
Run the same prompt on each set until the full list is complete.

Final Thoughts

Prompting is an art as much as a science. The more precise your recipe, the better the AI's cake. By giving clear instructions, examples, and manageable tasks, you'll get results that are accurate, useful, and tailored to your needs.

Prompt Recipes: How to Write Instructions AI Can Actually Follow