Write Better AI Prompts: A Simple Framework That Works Everywhere

The difference between a useless AI answer and a genuinely good one is almost never the model. It's the prompt. Vague question in, vague answer out. The fix isn't memorizing tricks — it's a simple, repeatable structure you can apply to any tool. Here it is.

Why most prompts fail

A typical prompt looks like: "Write me a blog post about productivity." The AI has no idea who you are, who it's writing for, how long it should be, or what "good" looks like to you — so it produces the blandest, most generic thing that technically matches. Garbage in, average out.

Good prompting is just giving the model the context a competent human would need to do the task well. That's the whole secret, and it fits in five parts.

The five-part framework

Think Role, Task, Context, Format, Constraints. You don't always need all five, but reaching for them turns weak prompts into strong ones.

1. Role

Tell the AI who to be. This sets vocabulary, depth, and perspective.

"You are an experienced technical writer who explains software to non-technical readers."

A role anchors the whole response. "Explain like a patient teacher" and "explain like a terse senior engineer" produce very different — and differently useful — answers.

2. Task

State exactly what you want done, with a strong verb. Not "help me with my email" but "Rewrite this email to be more concise and friendly." Specificity here is most of the battle.

3. Context

This is the part people skip, and it's the most valuable. Give the background a human would need: who the audience is, what you've already tried, what the goal is, relevant facts.

"This is for first-time users who've never used a spreadsheet. They get intimidated by jargon."

Context is the single biggest lever on output quality. When an answer disappoints, missing context is usually why.

4. Format

Describe the shape of the answer you want: a bulleted list, a table, three options, a 200-word paragraph, a step-by-step guide. If you don't specify, you get the model's default, which is rarely what you pictured.

"Give me five options as a numbered list, each one sentence."

5. Constraints

Set the boundaries: length, tone, what to avoid, what to include.

"Keep it under 150 words. No jargon. Don't use the word 'leverage.'"

Constraints are how you stop the model from rambling and steer it away from its worst habits.

Putting it together

Compare the weak version with the framework version.

Weak: "Write a product description for my app."

Framework:

"You are a conversion-focused copywriter (role). Write a product description for a habit-tracking app (task). It's aimed at busy professionals who've tried and abandoned other habit apps because they were too complex (context). Format it as a punchy headline plus three short benefit bullets (format). Keep it under 80 words, warm but not salesy, and don't make medical claims (constraints)."

The second prompt gets you something usable on the first try. The first gets you three more rounds of "no, more like…".

Iterate, don't restart

Even a great prompt rarely nails it first time, and that's fine. Treat it as a conversation:

• "Make it shorter."
• "The second option is closest — give me five more like that."
• "Too formal. Loosen it up."

Steering an answer toward what you want is faster than crafting the perfect opening prompt. The framework gets you 80% there; iteration covers the rest.

Save the ones that work

When a prompt produces consistently good results, save it. Build a small library of reliable prompts for your recurring tasks — summarizing, drafting, critiquing — and stop reinventing them. Keep it in your second brain or a snippet manager so it's one shortcut away. This is exactly the habit that separates people who get real value from AI from people who find it "hit or miss," a theme we dig into in using ChatGPT for work.

It transfers everywhere

The best part: this framework isn't tied to one tool. Role, Task, Context, Format, Constraints works in ChatGPT, Claude, Gemini, and whatever launches next month, because it's not a trick — it's just clear communication. Master it once and every AI tool you touch gets noticeably more useful.