Introduction
A good prompt does more than ask an AI model a question. It provides the model with the direction, context, constraints, and expected output needed to complete a task.
Prompt engineering techniques help users make their instructions more structured and predictable.
These techniques are useful for:
- AI chatbots
- Content generation
- Research
- Coding
- Business automation
- AI agents
- Data analysis
In Part 1, we covered the fundamentals of prompt structure. Now we will explore practical techniques for creating and improving prompts.
Prompt Engineering Techniques
There is no single prompt format that works for every task.
Different situations require different techniques.
Common approaches include:
- Role prompting
- Instruction prompting
- Contextual prompting
- Structured prompting
- Iterative prompting
- Prompt templates
- Example-based prompting
The best technique depends on the task and the AI system being used.
1. Role Prompting
What Is Role Prompting?
Role prompting gives the AI a defined perspective or responsibility for a particular task.
Example:
Act as an SEO content strategist. Analyze this article for search intent and suggest improvements.
The role provides additional context about the task.
When to Use Role Prompting
Role prompting can be useful when you want the response to follow a particular professional perspective.
Examples:
Act as a Python tutor and explain this code to a beginner.
Act as a marketing analyst and evaluate this campaign.
Act as a technical editor and identify unclear sections.
However, the role should not replace specific instructions.
A good prompt combines the role with a clear task.
2. Instruction Prompting
Instruction prompting directly tells the AI what action to perform.
Examples:
- Summarize this document.
- Compare these products.
- Rewrite this paragraph.
- Extract the key findings.
- Create a step-by-step guide.
A strong instruction should answer:
What should the AI do?
Weak Instruction
SEO.
This gives almost no direction.
Better Instruction
Analyze this article for on-page SEO and identify the five most important improvements.
The second prompt clearly defines the task.
3. Contextual Prompting
Contextual prompting provides relevant background information before asking the AI to perform a task.
Example:
I am writing a beginner-level article for students who have no programming experience. Explain APIs using simple language and practical examples.
The context tells the model:
- Who the audience is
- What level to use
- What style is appropriate
Why Context Matters
The same task can require different answers depending on the situation.
For example:
Explain artificial intelligence.
is different from:
Explain artificial intelligence to a 9th-class student using everyday examples.
The second prompt provides a much clearer context.
4. Structured Prompting
Structured prompts organize instructions into clearly defined sections.
For example:
Role: SEO writer Task: Write a beginner guide about technical SEO. Audience: New website owners Requirements: Use simple English. Include practical examples. Use clear headings. Output: Introduction + 5 sections + FAQ
This makes complex instructions easier to understand and modify.
Why Structured Prompts Are Useful
Structured prompts can help when a task has multiple requirements.
They are particularly useful for:
- Content creation
- Research
- Data extraction
- Coding
- Business workflows
- Repeated tasks
5. Iterative Prompting
Prompt engineering often works best as an iterative process.
Instead of expecting the first prompt to produce a perfect result:
Prompt ↓ Response ↓ Review ↓ Identify Problems ↓ Improve Prompt ↓ New Response
This allows you to refine the output.
Example of Iterative Prompting
First Prompt
Write a product description for a mattress topper.
Result
The description may be too generic.
Improved Prompt
Write a product description for a 10 cm mattress topper. Highlight comfort, pressure relief, washable cover, and UK bedroom use.
Further Improvement
Write a 200-word product description for a 10 cm mattress topper. Use a premium but natural tone. Highlight comfort, pressure relief, washable cover, and suitability for UK bedrooms. Avoid exaggerated medical claims.
Each revision adds useful constraints.
6. Prompt Templates
What Is a Prompt Template?
A prompt template is a reusable prompt structure that can be filled with different information.
Example:
Analyze the following [CONTENT TYPE] for [GOAL]. Target audience: [AUDIENCE] Requirements: [REQUIREMENTS] Output format: [FORMAT]
The same structure can be reused for many tasks.
Why Prompt Templates Are Useful
Templates are valuable when performing repetitive tasks.
For example, an e-commerce business could use one template for hundreds of product descriptions.
A content team could use another template for article optimization.
7. Example-Based Prompting
Providing examples can show the AI exactly what type of output you expect.
For example:
Input: Apple Output: Fruit, sweet, commonly eaten fresh. Input: Carrot Output: Vegetable, crunchy, commonly used in salads. Now process: Banana
The examples demonstrate the desired pattern.
This approach can be particularly useful for:
- Classification
- Formatting
- Data transformation
- Structured responses
Zero-Shot, One-Shot and Few-Shot Prompting
Example-based prompting leads to several important concepts.
Zero-Shot
No examples are provided.
Instruction ↓ AI Response
One-Shot
One example is provided.
Instruction + One Example ↓ AI Response
Few-Shot
Several examples are provided.
Instruction + Multiple Examples ↓ AI Response
These techniques will be explored in greater depth in a dedicated future article.
Prompt Templates for Different Tasks
A template can be adapted for different use cases.
Summarization Template
Summarize the following [DOCUMENT TYPE] for [AUDIENCE]. Highlight the main findings and important conclusions. Use [FORMAT].
Comparison Template
Compare [ITEM A] and [ITEM B] based on [CRITERIA]. Present the result in a table.
Content Template
Write a [WORD COUNT]-word article about [TOPIC] for [AUDIENCE]. Use [TONE] and include [SECTIONS].
Common Prompt Engineering Mistakes
Even when users understand prompting, they can make mistakes that reduce output quality.
Mistake 1: Being Too Vague
Bad:
Make it better.
Better:
Rewrite this paragraph to improve clarity and readability while preserving its original meaning.
Mistake 2: Giving Conflicting Instructions
For example:
Make the answer extremely detailed but keep it under 50 words.
Conflicting requirements can make the desired result unclear.
Instead, prioritize the most important requirement.
Mistake 3: Providing Too Much Irrelevant Context
More information does not automatically produce better results.
Unnecessary context can make the prompt harder to manage.
Use:
Relevant information > excessive information
Mistake 4: Not Defining the Output
If you need a table, say so.
If you need bullet points, specify it.
If you need JSON, define the required structure.
Example:
Return the results as a table with columns for topic, difficulty, and recommendation.
Mistake 5: Asking for Multiple Unrelated Tasks
A prompt containing many unrelated tasks can become difficult to manage.
Instead of:
Write an article, analyze my website, create a marketing plan, translate it, and write five social posts.
Consider separating the tasks or clearly organizing them into stages.
Mistake 6: Assuming AI Is Always Correct
A well-written prompt does not guarantee factual accuracy.
AI systems can still produce:
- Incorrect information
- Outdated information
- Unsupported claims
- Misinterpretations
Important information should be verified.
Mistake 7: Using Unclear References
Avoid prompts such as:
Fix this and make it like that.
Instead, identify exactly what "this" and "that" mean.
Clear references reduce ambiguity.
Practical Prompt Engineering Example
Suppose you want an AI to write an SEO article.
Basic Prompt
Write an article about SEO.
Improved Prompt
Write a beginner-friendly article explaining SEO.
Structured Prompt
Role: Act as an SEO content writer. Task: Write a beginner guide to SEO. Audience: New website owners. Requirements: Explain keywords, on-page SEO, technical SEO, and backlinks. Use simple English. Include practical examples. Output: Use an introduction, H2 headings, bullet points, FAQ, and conclusion.
The structured version provides much stronger direction.
Prompt Engineering for AI Agents
Prompt engineering becomes particularly important when AI models are used inside AI agents.
An agent may need instructions about:
- Goals
- Available tools
- Rules
- Constraints
- Expected actions
A simplified workflow:
Agent Goal ↓ System Instructions ↓ Context ↓ Available Tools ↓ LLM ↓ Action
Prompt Engineering and Structured Outputs
For applications, it can be useful to specify an exact output structure.
For example:
{ "topic": "...", "summary": "...", "priority": "..." }
Structured outputs make AI responses easier for software systems to process.
Future of Prompt Engineering
As AI models become more capable, prompt engineering may evolve.
Future AI systems may require less detailed prompting for simple tasks.
However, complex applications will still need clear specifications involving:
- Goals
- Context
- Constraints
- Tools
- Data
- Output requirements
Prompt engineering may increasingly become part of broader AI system design rather than simply writing clever sentences.
Frequently Asked Questions
What is the best prompt engineering technique?
There is no universal best technique. The right approach depends on the task. Clear instructions, relevant context, and well-defined outputs are useful foundations.
Does role prompting always improve AI responses?
Not necessarily. A role can provide useful context, but the task and requirements still need to be clearly specified.
Are longer prompts better?
No. A longer prompt is not automatically better. Relevant and well-organized information is generally more useful than unnecessary detail.
What are prompt templates?
Prompt templates are reusable prompt structures containing placeholders that can be filled with different information.
Can prompt engineering eliminate AI hallucinations?
No. Better prompts can improve clarity and consistency, but they cannot guarantee factual accuracy.
Is prompt engineering only for developers?
No. Anyone using AI tools can benefit from basic prompt engineering skills.
Conclusion
Prompt engineering is fundamentally about clear communication with AI systems.
The most useful techniques include:
- Role prompting
- Instruction prompting
- Contextual prompting
- Structured prompting
- Iterative prompting
- Prompt templates
- Example-based prompting
The goal is not to make prompts unnecessarily complicated.
Instead, a strong prompt should clearly communicate:
What to do + relevant context + requirements + desired output.