The 5-part formula for a great prompt
The fastest way to get expert results from any AI is to give it five things: a Role (who it should act as), Context (the situation and audience), the Task (exactly what you want), Constraints (rules, length, tone, what to avoid), and the Output Format (how the answer should be structured). A one-line question gives the model nothing to work with, so it defaults to generic. A prompt with these five parts gives it everything it needs to produce something specific and usable on the first try. You do not need clever tricks — you need structure.
1. Assign a role
Start by telling the model who to be: a senior copywriter, a tax accountant, a startup advisor, a patient tutor. A role instantly loads the right vocabulary, assumptions, and standards. Compare a bare request like write a sales email to act as a direct-response copywriter who has written for SaaS companies for ten years. The second produces sharper, more professional output because the model now has a perspective to write from. Be specific about seniority and specialty — expert beats generic every time.
2. Give context
The model cannot read your mind, so spell out the situation: who the audience is, what you are trying to achieve, what has already been tried, and any background that matters. For a sales email, context means the product, the customer, the goal of the email, and the desired tone. Context is the single biggest lever on quality — most disappointing AI output traces back to a prompt that assumed the model knew things it could not possibly know.
3. State the task precisely
Say exactly what you want done, as one clear instruction. Vague verbs like help me or improve this leave too much room. Replace them with specifics: write a 4-email sequence, rewrite this paragraph to be 30 percent shorter, list the 10 biggest risks ranked by severity. The more precisely you define the deliverable, the less the model has to guess — and guessing is where generic output comes from.
4. Add constraints
Constraints turn a decent answer into the right answer: length (under 150 words), tone (confident, not salesy), audience reading level, things to avoid (no jargon, no clichés), and any must-include points. Constraints are also how you stop the model from padding with filler. A good rule: every time the AI gives you something almost-right, the fix is usually a constraint you forgot to state.
5. Specify the output format
Tell the model how to structure the answer: a table, numbered steps, a JSON object, a two-column comparison, bullet points with bold headers. Format instructions make the output immediately usable and easy to scan, and they are essential when you plan to copy the result into another tool. Add reusable placeholders in brackets — like PRODUCT or AUDIENCE — so you can save the prompt as a template and reuse it forever. That last step is what turns a one-off prompt into a permanent productivity asset.
Frequently Asked Questions
What is the best structure for an AI prompt?
The most reliable structure has five parts: Role (who the AI should act as), Context (the situation and audience), Task (exactly what you want), Constraints (rules, length, tone, what to avoid), and Output Format (how to structure the answer). Including all five consistently produces expert-quality results on the first try.
Why are my AI prompts giving generic answers?
Almost always because the prompt lacks context and constraints. The model fills gaps with generic defaults, so if you do not specify the audience, goal, tone, and format, you get average output. Add a role, the background, and a precise deliverable and the quality jumps immediately.
Do I need different prompts for ChatGPT, Claude, and Gemini?
No. A well-structured prompt transfers across all major models with little or no change. The five-part structure matters far more than the specific model. Minor tuning can help, but you can keep one prompt library and use it everywhere.
What are placeholders in a prompt?
Placeholders are bracketed variables that mark the parts you swap each time, such as PRODUCT, AUDIENCE, or TONE. They turn a one-off prompt into a reusable template — you write it once and just fill in the blanks for every future use.
How long should an AI prompt be?
As long as it needs to be to include role, context, task, constraints, and format — and no longer. For simple tasks that may be a few sentences; for complex deliverables it may be several short paragraphs. Clarity matters more than length; cut anything that does not change the output.