Guide · Hazel
How to Write Better AI Prompts in 2026
Short answer: The fastest way to write better AI prompts is to structure them: give the AI a Role (the expert to act as), Context (the specifics of your situation), a Task (exactly what to produce), Constraints (rules and limits), and an Output format. This five-part structure turns vague requests into consistent, expert-level results in ChatGPT, Claude or Gemini.
The Role–Context–Task–Constraints–Output framework
- Role. Tell the AI who to be: "You are a senior conversion copywriter." This sets the standard for everything that follows.
- Context. Give the real specifics — audience, product, situation, what you've tried. Vague context is the #1 cause of generic answers.
- Task. State exactly what to produce, step by step if needed. Ambiguity gets interpreted in the laziest way.
- Constraints. Add the rules: length, tone, what to avoid, what to refuse. Constraints are where quality lives.
- Output format. Say how you want it back — a table, a numbered list, a 700-word memo. You'll get something usable instead of a wall of text.
Before and after
Weak: "Write me a landing page." Better: "Role: senior conversion copywriter. Context: I sell a €30 one-time AI prompt system to non-technical founders. Task: write a hero headline, subhead and primary CTA, 3 variations. Constraints: no hype, benefit-led, under 12 words per headline. Output: a numbered list." Same model, far better result.
When to use a skill instead
If you find yourself pasting the same Role and Constraints into every chat, turn it into an installable skill — load it once and every prompt inherits it. That's the difference between writing better prompts and building a system. (See prompts vs skills.)
Why filled examples beat blank templates
Seeing a fully worked prompt — with the brackets already filled in once — teaches the pattern far faster than a blank template. That's how Hazel's 323 prompts are written: each ships with a real example you adapt. Browse what's inside or get everything in The Vault & Xtras for €30.
Hazel is an educational toolkit; your results depend on your own effort.
Frequently asked questions
How do I write a good AI prompt?
Structure it: Role (the expert to act as), Context (your specifics), Task (exactly what to produce), Constraints (rules and limits), and Output format. This turns vague requests into consistent, expert-level results.
What is the best prompt structure in 2026?
Role–Context–Task–Constraints–Output is a simple, repeatable structure that works across ChatGPT, Claude and Gemini. It forces you to give the model what it needs to answer well.
Why do my AI prompts give generic answers?
Usually missing context and constraints. Add who the answer is for, the real specifics of your situation, and the rules/format you want — generic in, generic out.
When should I use a skill instead of a prompt?
When you repeat the same role and rules constantly. Turn them into an installable skill you load once, so every later prompt inherits that expertise automatically.