A Simple Structure for Writing Prompts

This article is practical and introductory in nature. It is intended for users who are new to writing prompts for generative AI systems.

In the previous section, we explained that generative AI systems do not execute prompts directly.

Instead, prompts are interpreted and compressed internally before generation.

Conceptually:

A (prompt)
→ A′ (internal interpretation)
→ B′ (generated output)

Because of this process, the way a prompt is structured can strongly influence how the AI interprets it.

When beginners start writing prompts for generative AI, they often feel confused.

Many prompt guides immediately introduce abstract structures such as:

Theme
Rules
Materials
Output

For beginners, this can easily feel like: “What is this supposed to mean?”

A simpler approach is to start with a structure that matches how people naturally think.

Step 1 — Start with Five Simple Elements

For beginners, prompts are easier to understand when they are broken into five parts:

Theme
Rules
Materials
Process
Output

These correspond to very natural questions.

Theme — What is the topic?

Rules — What conditions should be followed?

Materials — What information or documents should be referenced?

Process — How should the answer be developed?

Output — What form should the final answer take?

This structure mirrors the way humans usually organize tasks.

Theme (What is the topic?)
        ↓
Rules (Conditions to follow)
        ↓
Materials (Reference information)
        ↓
Process (Steps to follow)
        ↓
Output (Final format)

The purpose of this structure is not to make prompts longer. It is to reduce ambiguity.

By separating the topic, conditions, reference materials, reasoning process, and final format, users can make it easier for the AI to preserve the original intent of the prompt.

Because it follows a natural mental flow, beginners can understand it quickly.

Step 2 — Understand the Role of Each Part

Each element in the structure plays a different role.

Understanding these roles helps prevent common mistakes when writing prompts.

Theme

The Theme defines what the prompt is about.

In some prompts, the theme may also include the role the AI should take or the goal the user wants to achieve.

Example:

Theme:
Explain how generative AI works for beginners.

Without a clear theme, the AI may interpret the task too broadly.

Rules

Rules define the conditions that the AI should follow.

Example:

Rules:
Use simple language
Avoid technical jargon

Rules guide how the AI should approach the task. They act as behavioral constraints on the response.

Materials

Materials contain reference information the AI should use.

Example:

Materials:
Notes about generative AI behavior
An article about diffusion models

Beginners often make the mistake of mixing instructions with materials. However, these are different things.

Rules = instructions that guide behavior
Materials = reference information the AI may draw on

Materials provide context and information. They do not tell the AI how to behave.

Process

The Process describes the steps the AI should follow before producing the final answer.

Example:

Process:
1. Explain the basic idea
2. Provide a simple example
3. Summarize the key point

The process controls the path the answer should take, not just its content.

Output

The Output defines the format of the final result.

Example:

Output:
Use bullet points
Use a Markdown table if comparison is needed
Keep the answer under three paragraphs

Output instructions control how the answer is presented. They do not define what the answer contains.

Step 3 — Simplifying the Structure

Once you become comfortable with prompts, you may notice that some elements overlap.

In particular, the Process is often simply a type of Rule.

For example:

Rules:
Explain the concept first
Provide an example
Summarize the key idea

In other words: Process → part of Rules

Because of this, the structure can often be simplified.

The Final Four-Part Structure

After simplification, prompts can usually be organized into four elements.

Element	Role
Theme	What the prompt is about
Rules	How the AI should respond
Materials	What information it should use
Output	How the answer should appear

Why This Order Matters

If beginners are shown the four-part structure first, it can feel abstract and confusing.

That is why it is easier to introduce prompts in two stages.

First:

Theme → Rules → Materials → Process → Output

Then simplify:

Theme → Rules → Materials → Output

This progression follows a natural learning path:

Concrete idea
        ↓
Organized structure
        ↓
Simplified model

By starting with a structure that matches how people naturally think, prompt design becomes much easier to understand.

Where This Structure Applies

This structure can be used with many current generative AI systems, especially text-based conversational models.

However, different systems may require different emphasis depending on their purpose. Search-based systems may require clearer source instructions, while image generation systems often use a more visual prompt structure.

Summary

Effective prompts often share several common characteristics.

In practice, prompts tend to work best when they follow three basic principles:

information is clearly organized
instructions are structured
requests are concise and specific

These properties make prompts easier for the AI to interpret when it reconstructs their meaning internally.

Because generative AI systems internally reinterpret prompts before generating outputs, well-structured and concise prompts are generally more stable and easier for the model to follow.

In conversational systems, interactions can also be understood as a sequence of prompt–response cycles.

Conceptually, a conversation can be seen as:

Prompt → Response → Prompt → Response → ...

Each new message becomes part of the context that the AI interprets when generating the next response.

Understanding prompts in this way helps explain why:

clear instructions improve reliability
long, unstructured prompts may become unstable
long conversations can gradually change how the AI interprets a task

By thinking of prompts as structured information given to a probabilistic system, users can design prompts more effectively and avoid many common frustrations.

Next Section

This structure works well for many text-based prompts.

However, image generation prompts often follow a slightly different pattern. They usually place more emphasis on subject, style, composition, visual details, and constraints.

In the next section, we will examine how prompts are commonly written for image generation tasks.