The Anatomy of a Good Prompt: Four Elements That Change Everything

Why Most Prompts Fall Short

Most people write prompts the way they send a text message: short, informal, assuming the other party understands the context. This works fine between people who share context. It does not work well with AI, which knows nothing about you or your situation unless you tell it.

The solution is not to write longer prompts for their own sake. It is to include the right information. There are four elements that, when included, consistently produce better output: the task, the context, the format, and examples. Learn to include all four and your results will improve dramatically.

Element 1: The Task

The task is what you want the AI to do. This sounds obvious, but most people are less specific than they think.

"Write about our product" is a task. But it leaves almost everything open. Is it an ad, a blog post, a product description, an internal memo? How long? What goal?

"Write a 150-word product description for [product name] that highlights its three main benefits and ends with a purchase CTA" is a much better task. It specifies the format (product description), the length (150 words), the content (three benefits), and the goal (purchase CTA).

Practice: Before writing your next prompt, ask yourself: have I told the AI exactly what kind of output I want, how long it should be, and what its purpose is?

Element 2: The Context

Context is everything the AI needs to know to do the task well. This includes information about you, your audience, your product or topic, and any constraints or preferences that apply.

Without context, the AI writes for a generic reader with no particular knowledge of your situation. With context, it writes specifically for you.

Examples of useful context:

• "My audience is small business owners in the UK with no accounting background."
• "Our brand tone is direct and practical. We avoid corporate jargon."
• "This is for an internal presentation, not for customers."
• "The product costs $299 and our main competitor charges $499."

You do not need to include everything. Include the context that would change what a capable human writer would produce if they knew it.

Practice: Ask yourself: what would a smart human writer need to know about my situation to do this task well? Tell the AI that.

Element 3: The Format

Format tells the AI how to structure the output. If you do not specify a format, the AI picks one, and it may not be what you wanted.

Format instructions can specify:

• Length ("under 100 words", "three to four paragraphs", "exactly 5 bullet points")
• Structure ("use headers", "write as a numbered list", "return JSON")
• Tone ("formal", "conversational", "dry and technical")
• What to include or exclude ("do not use bullet points", "include a subheading for each section", "no introduction paragraph")

Being specific about format saves you editing time. If you need a three-paragraph response and the AI writes eight paragraphs, you have to cut it down. If you said "three paragraphs" in the prompt, it usually delivers exactly that.

Practice: Before you send a prompt, decide what the output should look like and add those specifications. Even one or two format constraints make a meaningful difference.

Element 4: Examples

Examples are the most powerful element of a prompt, and the one most people skip. When you show the AI an example of the output you want, it learns from that example far more effectively than from a description.

Suppose you want the AI to write in a specific style. You could describe that style in words, or you could show it three sentences written in that style and say "write more like this." The second approach almost always works better.

Examples can be:

• A sample output: "Here is an example of the kind of response I want: [example]"
• A before-and-after: "Bad version: [example]. Good version: [example]. Now do the same for:"
• Your own past work: "Here are three emails I have written that have my voice: [examples]. Write a new email in the same voice."

You will learn more about examples in the few-shot prompting tutorial coming up in this course.

Putting It Together: A Template

Here is a simple template you can apply to almost any prompting task:

[TASK]
Write / summarize / analyze / draft [specific output type].

[CONTEXT]
Audience: [who will read this]
Tone: [how it should sound]
Background: [key information the AI needs]

[FORMAT]
Length: [word count or structure]
Structure: [headers / bullet points / paragraphs / JSON / etc.]
Include: [specific elements to include]
Exclude: [things to leave out]

[EXAMPLES] (optional but powerful)
Here is an example of what I am looking for:
[your example]

You do not need to use all four elements every time. A simple task with a clear audience might only need task and format. A complex task requiring a specific style will benefit from all four. The habit to build is checking for each element before you send a prompt and asking whether including it would improve the result.

A Before and After

Before: Write a LinkedIn post about our new feature.

After:

Write a LinkedIn post about our new feature: AI-powered invoice processing.

Audience: Finance directors and CFOs at mid-sized companies.
Tone: Professional but not stuffy. Conversational and direct.
Background: The feature saves an average of 4 hours per week per finance team member.
We want to communicate real business value, not technical details.

Format:
- Opening hook (one strong sentence)
- 3 short paragraphs covering the problem, the solution, and the result
- One closing line with a soft CTA (no hard sell)
- Under 200 words total
- No hashtags

The second prompt takes 30 seconds longer to write. The output requires almost no editing. That trade-off gets better the more you do it.