How to Write Better AI Prompts

A practical framework for writing AI prompts that get useful results — using role, task, context, constraints, and format.

Quick Answer

Most prompts get generic results because they’re too vague. Add a clear task, relevant context, and the output format you want. Use the RTCCF framework: Role, Task, Context, Constraints, Format. Iterate from there.

Why Vague Prompts Get Vague Results

AI responds to the quality of the input you give it. When you write “write me an email,” the AI has almost no information to work with — it doesn’t know who it’s for, what tone fits, what problem it’s solving, or how long it should be. So it guesses. And the guess is always generic.

When you give the AI specific context, it can match the tone, fit the audience, stay on topic, and produce something that actually resembles what you needed.

The gap between a useful AI output and a useless one is almost always in the prompt, not the tool.

The RTCCF Framework

Five elements that make prompts work:

R — Role

Tell the AI who it should be for this task. This sets the expertise level and perspective.

Weak: Write a product description.

With role: You are an experienced ecommerce copywriter who specializes in making technical products sound approachable. Write a product description for…

You can also use role to set the relationship: “You are a senior editor reviewing my draft” produces different feedback than “You are a supportive writing coach.”

T — Task

State clearly what you want done. Use a specific verb: write, rewrite, summarize, analyze, compare, list, explain, translate, create.

Weak: Help me with my presentation.

With task: Write an outline for a 10-minute presentation about our Q2 marketing results. Include an opening hook, 3 main sections, and a clear recommendation at the end.

C — Context

Give the AI the background it needs to do the task well. This includes:

• Who the audience is
• What the purpose is
• Any relevant background information
• What platform or format this is for

Weak: Write a LinkedIn post about our new product launch.

With context: Write a LinkedIn post announcing the launch of our new project management tool aimed at freelancers. Our audience is independent designers and consultants. The main benefit is that it connects client communication and invoicing in one place. We want to sound professional but not corporate.

C — Constraints

Tell the AI what to avoid, limit, or stay within.

• Word count or length limits
• Tone restrictions (“avoid formal language”)
• Things not to include (“don’t mention pricing”)
• Style rules (“use short sentences, no bullet lists”)
• Things the audience might find off-putting

Constraints are underused by beginners. They often matter as much as what you’re asking for.

F — Format

Specify how you want the output structured.

• Bullet list or numbered list
• Short paragraph
• Table
• Sections with headers
• A specific template format
• Code block

If you don’t specify format, AI will guess — and it often guesses wrong.

Before and After Examples

Email

Before:

Write a follow-up email.

After:

Write a follow-up email to a potential client who attended our product demo last Tuesday but hasn't responded to our proposal yet.

Tone: friendly but professional
Length: under 100 words
Goal: get a reply, not a commitment
Avoid: pressure tactics, assumptions about why they haven't replied
End with: a low-friction question they can easily answer

Blog post intro

Before:

Write a blog post intro about remote work.

After:

Write a 100-word opening paragraph for a blog post called "Why Remote Work Is Harder Than It Looks (And How to Fix It)". 

Audience: managers who are new to managing remote teams
Tone: honest and direct, not preachy
Start with: a specific observation that will resonate, not a statistic

Summary

Before:

Summarize this.

After:

Summarize this document for a busy executive who has 2 minutes to read it. Focus on: the main recommendation, the key supporting reasons, and any risks mentioned. Skip background context. Use bullet points. Total length: under 120 words.

[paste document]

The Refinement Loop

You don’t need a perfect prompt the first time. Write a reasonable prompt, read the output, then refine.

Common refinement requests:

Make this shorter — cut it to 80 words

The tone is too formal. Rewrite it to sound more like a real person talking

Keep the structure but change the examples to something more relevant to [your industry]

This is good but the opening is weak. Give me 5 alternative opening sentences

What did you leave out that might be important?

The refinement loop is one of the most productive habits you can build. The best results often come from the third prompt, not the first.

Prompt Templates You Can Reuse

The universal task brief

Role: [who you want the AI to be]
Task: [specific verb + what to produce]
Audience: [who will read or use this]
Context: [relevant background]
Constraints: [length, tone, things to avoid]
Format: [how to structure the output]

Edit anything

Edit this for clarity, conciseness, and tone. Flag anything that sounds unclear, overly formal, or out of place.

[paste text]

Brainstorm with guardrails

Give me 10 ideas for [goal].

Audience: [who this is for]
Constraint: [what to focus on or avoid]
Format: One line per idea with a brief reason why it works

Explain a concept

Explain [concept] as if I know [related thing] but have never encountered this before. Use a concrete example. Keep the explanation under 200 words.

Improve a specific part

This sentence is weak: "[paste sentence]"

Rewrite it to be more specific and direct. Give me 3 versions.

Common Prompting Mistakes

Telling AI what you don’t want without saying what you do want

“Don’t make it too salesy” is vague. “Use a helpful, informative tone — no calls to action, no urgency language” is specific.

Forgetting to say who the audience is

The same topic needs completely different language for a beginner vs an expert. Always specify.

Not specifying format

If you want a table, ask for a table. If you want three short paragraphs, say so. Format affects usability more than most people realize.

Treating AI output as the final version

AI output is a strong starting point, not a finished product. Plan to edit. Your own judgment, voice, and knowledge of context should shape the final result.

Using one massive prompt instead of multiple focused ones

If you’re asking for something complex, break it into steps. First ask for an outline. Review it. Then ask for the first section. Review it. This produces better results than one giant prompt that tries to do everything at once.

Final Takeaway

Better prompts come from more specificity, not more effort. Add a clear role, a specific task, relevant context, useful constraints, and the output format you need. Then iterate from the first draft.

The single most effective habit: after getting an output, ask “What would make this 30% better?” and then ask AI to make that change. Repeat once more. That three-step process almost always produces something genuinely useful.