What Is the CoSTAR Framework?
CoSTAR stands for Context, Objective, Style, Tone, Audience, and Response. Each letter represents a piece of information that helps the AI model understand not just what you want, but how you want it delivered and who it's for.
Developed as a structured prompting method for large language models, CoSTAR has become one of the most widely used frameworks in prompt engineering because it works with minimal effort and produces dramatically more consistent results than unstructured prompts.
| Letter | What it means | Example |
|---|---|---|
| C | Context — background information | "I run a SaaS startup with 50 customers" |
| O | Objective — the actual task | "Write a customer win-back email" |
| S | Style — how it should be written | "Direct, no fluff, like a Basecamp email" |
| T | Tone — emotional register | "Warm but not apologetic" |
| A | Audience — who will read it | "Customers who churned 30-60 days ago" |
| R | Response — format and length | "Under 120 words, subject line included" |
Why Most Prompts Fail
The average AI prompt looks like this: "Write me a marketing email."
The AI has no idea who the email is for, what product it's about, what tone to use, or how long it should be. So it produces something generic — a template that could have been written for anyone. You then have to rewrite most of it anyway.
"Write me a marketing email about my new course."
"Context: I teach prompt engineering online. Objective: Write a launch email for my new beginner course. Style: Conversational, like Paul Graham writing to friends. Tone: Excited but not salesy. Audience: Email subscribers who know me but haven't bought yet. Response: Under 200 words, subject line included, one CTA only."
The second prompt takes 30 extra seconds to write. The output requires zero editing.
How to Write a CoSTAR Prompt Step by Step
Step 1 — Context
Give the AI the background it needs to understand your situation. This is not the task — it's the setup. Think of it as the first paragraph of a brief you'd hand to a freelancer.
Good context includes: your role, your company or product, relevant constraints, and any information the AI couldn't know otherwise.
Step 2 — Objective
State exactly what you want the AI to produce. Be specific. "Write copy" is weak. "Write three subject line options for a win-back email campaign" is strong.
Step 3 — Style
This is the most underused part of CoSTAR and the one that makes the biggest difference. Style tells the AI how the writing should feel — not the topic, but the approach. Reference a real writer, publication, or brand voice if you can. "Write like The Economist" or "write like a Stripe blog post" gives the AI a concrete target.
Pro tip: Instead of describing the style, show it. Paste 2-3 sentences written in the style you want and say "match this tone." A concrete example outperforms any description.
Step 4 — Tone
Tone is the emotional register — how the writing should make the reader feel. Confident, empathetic, urgent, playful, authoritative. Be specific. "Professional" means nothing. "Confident without being aggressive, like a founder who knows their product is good" means something.
Step 5 — Audience
Describe who will read or use the output. Age, role, knowledge level, context. An explanation written for a senior engineer should read completely differently from one written for a first-time user. Without this, the AI aims at nobody.
Step 6 — Response
Specify the format, length, and structure of the output. Do you want bullet points or prose? Headers or none? A specific word count? JSON? Markdown? The more precise you are here, the more usable the output is immediately.
A Full CoSTAR Example
Here's a complete CoSTAR prompt for a LinkedIn post:
CoSTAR vs Other Frameworks
CoSTAR isn't the only prompt framework — it's just one of the most consistently reliable. Here's how it compares to two other popular options:
| Framework | Best for | When to use |
|---|---|---|
| CoSTAR | General content, emails, copy | When you need polished, audience-aware output |
| RISEN | Step-by-step tasks with constraints | When the AI needs to follow a specific process |
| Chain-of-Thought | Complex reasoning, analysis | When accuracy matters more than style |
Common CoSTAR Mistakes
- Skipping Style and Tone. These two fields do more work than any other part of the framework. Without them, every output sounds the same.
- Being vague about the Audience. "General audience" is not an audience. Name a specific person or persona.
- Ignoring Response format. If you don't specify a format, the AI picks one — and it's often not what you wanted.
- Making Context too long. Aim for 1-3 sentences. A 500-word context block dilutes the signal.
Key insight: The negative constraints in your Response field do more work than the positive ones. "No bullet points, no exclamation marks, under 150 words" produces better output than "clear, concise, and well-formatted."
Try It Right Now
The fastest way to internalize CoSTAR is to use it once on something real. Pick a task you actually need done — an email, a social post, a product description — and run it through the six fields before you send the prompt.
You'll notice the difference in the first output. The AI stops guessing and starts delivering.