Bottom line
How to get better results from AI models. Practical prompting techniques that work across GPT, Claude, and Gemini.
The Basics
Be Specific
Vague prompts = vague results. Instead of “Write code,” say “Write a TypeScript function that validates email addresses.”
Provide Context
Give background. “We’re building a SaaS app” is better than nothing.
Show Format
Tell the AI exactly how you want output: “Return JSON with fields: id, name, price.”
Advanced Techniques
Chain of Thought
Ask for step-by-step reasoning:
“Think step by step before answering.”
Few-Shot Examples
Show examples of desired output:
“Example: Input: 2+2 Output: 4”
Role Prompting
Assign a persona:
“You are a senior backend engineer with 15 years experience.”
Constraint Setting
Be explicit about limits:
“Provide a solution under 50 lines. Don’t use external libraries.”
Model-Specific Tips
GPT-5
- Great with system prompts
- Excels at following formats
- Good with chain-of-thought
Claude 4
- Best at long context tasks
- Excellent at code review
- Responds well to detailed instructions
Gemini
- Fast, use for simple tasks
- Good at multimodal (images + text)
Common Mistakes
- Not being specific enough
- Missing context
- No output format specified
- Asking too much at once
- Not iterating — first attempt usually not best