Learn prompt engineering from scratch — what it is, why it matters, 5 key techniques with practical examples, common mistakes to avoid, and the best tools to practise with.
There's a reason some people get genuinely useful, production-ready output from AI tools while others get generic, frustrating responses. The difference usually isn't the model — it's the prompt.
Prompt engineering is the practice of designing inputs to AI systems to reliably get the outputs you want. It sounds simple. In practice, it's a skill that takes deliberate effort to develop, and it's already becoming one of the most in-demand competencies in the AI era.
This guide will take you from zero to confident — covering what prompt engineering is, the five core techniques, ten practical before/after examples, the most common mistakes, and the tools you should be practising with.
A prompt is any input you give to a large language model (LLM) — a question, a command, some context, or a combination of all three.
Prompt engineering is the systematic practice of crafting those inputs to steer the model towards the response you actually need. It's part science (there are well-established techniques) and part art (context, nuance, and creativity matter a lot).
Here's the core insight: LLMs don't "understand" what you want. They predict the most statistically likely continuation of your text based on patterns in their training data. The way you phrase your prompt dramatically changes what patterns the model reaches for — and therefore what it produces.
A vague prompt gives you a vague answer. A well-structured prompt gives you something you can actually use.
Zero-shot means giving the model a task with no examples. You're relying entirely on what the model learned during training.
This works well for tasks the model has seen thousands of times — summarising, translating, basic Q&A. It struggles for niche, complex, or nuanced tasks.
Example:
Summarise the following article in three bullet points:
[article text here]
Zero-shot is your default. Use it when the task is clear and common. Upgrade to one of the techniques below when you're not getting what you need.
Few-shot means providing two to five examples of the input-output pattern you want before giving the model your actual task. You're showing it the format and tone you expect.
This is one of the most reliable techniques for getting consistent output formatting.
Example:
Classify the sentiment of the following customer reviews as Positive, Negative, or Neutral.
Review: "The delivery was fast and the product was exactly as described."
Sentiment: Positive
Review: "It broke after two days. Total waste of money."
Sentiment: Negative
Review: "It arrived on time. Does what it says."
Sentiment: Neutral
Review: "I've ordered three times now and every time has been a great experience."
Sentiment:
The model sees the pattern and continues it. You get consistent formatting without writing complex instructions.
Chain-of-thought prompting asks the model to reason through a problem step by step before giving a final answer. This dramatically improves performance on logic, maths, multi-step problems, and anything requiring reasoning.
The simplest version: add "Let's think step by step" to your prompt. That's it. It sounds almost too simple, but it consistently improves answer quality on complex tasks.
Example — without CoT:
A shop sells apples for £0.40 each and bananas for £0.25 each.
Emma buys 5 apples and 8 bananas. She pays with a £10 note.
How much change does she get?
Example — with CoT:
A shop sells apples for £0.40 each and bananas for £0.25 each.
Emma buys 5 apples and 8 bananas. She pays with a £10 note.
How much change does she get?
Think through this step by step before giving your final answer.
The second version is far less likely to give you a wrong answer. The model is forced to show its working, which catches arithmetic errors.
CoT is especially valuable when you need the model to explain its reasoning — for educational content, legal analysis, or any task where "trust me" isn't good enough.
Role prompting assigns a specific persona, expertise, or perspective to the model. Framing the model as an expert in a particular domain nudges it to draw on the relevant patterns from its training data — and changes the tone, vocabulary, and assumed knowledge level of the response.
Basic role prompt:
You are a senior software engineer with 15 years of experience in Python.
Review the following code and identify any bugs, performance issues, or non-Pythonic patterns:
[code here]
Role + audience prompt (even more powerful):
You are an experienced primary school teacher explaining concepts to 10-year-olds.
Explain how the internet works. Use a simple analogy and avoid technical jargon.
The role prompt does two things: it tells the model what to know and how to communicate it. Both matter.
One of the most underused techniques is explicitly specifying the format you want. Models will adapt to almost any output structure if you describe it clearly — JSON, Markdown tables, numbered lists, specific headings, constrained length.
Without formatting instructions:
What are the main differences between supervised and unsupervised learning?
Result: A prose paragraph of variable length and structure.
With formatting instructions:
What are the main differences between supervised and unsupervised learning?
Format your response as a Markdown table with three columns: Aspect, Supervised Learning, Unsupervised Learning.
Include at least 5 rows covering: Definition, Training Data, Goal, Common Algorithms, Example Use Cases.
Result: A clean, consistent table you can paste directly into documentation.
This technique is essential when you're automating tasks or building pipelines where downstream systems expect a specific format.
Before (weak):
Write an intro for a blog post about AI.
After (strong):
Write a compelling opening paragraph (80–100 words) for a blog post titled
"What Is Generative AI? A Beginner's Complete Guide."
The audience is working professionals with no technical background.
Start with a hook — a surprising stat or a relatable scenario — then introduce the topic.
Don't use the phrase "In today's digital age."
Before:
Explain neural networks.
After:
Explain how neural networks work to someone who understands basic maths (addition,
multiplication) but has no programming experience. Use a concrete analogy —
not "it's like the human brain." Keep it under 200 words and end with
one sentence on why this matters practically.
Before:
Summarise this document.
After:
You are an executive assistant summarising a report for a busy CEO.
Summarise the attached document in the following format:
**One-sentence summary:** (max 25 words)
**Key findings:** (3–5 bullet points)
**Recommended actions:** (if any, max 3)
**Risks or concerns:** (if any, max 3)
Before:
Fix my code.
After:
You are a senior Python developer. The following function is supposed to
return the most frequently occurring word in a list of strings, but it's
returning incorrect results for some inputs.
1. Identify the bug.
2. Explain why it causes the problem.
3. Provide the corrected code.
4. Suggest one test case that would catch this bug.
[code here]
Before:
Write an email to my client about the project delay.
After:
Write a professional email to a long-term client (B2B relationship, 3 years)
explaining that a software project delivery will be delayed by 2 weeks due to
unexpected technical issues in the third-party API we're integrating with.
Tone: apologetic but confident. Length: under 150 words.
End with a clear next step — a call scheduled for this week.
Do not include a subject line.
Before:
Write a tweet about our new course launch.
After:
Write 3 different tweets announcing the launch of a free AI course for beginners.
Each tweet must:
- Be under 240 characters
- Include a hook, the core benefit, and a call to action
- Have a different angle: one aspirational, one practical, one curiosity-driven
- End with the hashtags #LearnAI #AIEducation
Do not use emojis.
Before:
Analyse these customer reviews.
After:
You are a UX researcher analysing 50 customer reviews for a mobile banking app.
From the reviews below, identify:
1. Top 3 praised features (with frequency estimate)
2. Top 3 complained-about issues (with frequency estimate)
3. One surprising insight that isn't obvious
4. One specific product recommendation based on the data
Be specific. Quote the reviews where relevant.
[reviews here]
Before:
Teach me about transformers.
After:
I'm a software developer comfortable with Python and basic ML concepts
(I know what a neural network is). I want to understand the Transformer
architecture well enough to explain it to a colleague.
Teach me in this order:
1. The problem Transformers were designed to solve (why RNNs weren't enough)
2. The core idea of self-attention in plain language
3. A simplified walkthrough of the encoder-decoder structure
4. One concrete, real-world example of Transformers in action
Use analogies. Skip the maths for now.
Before:
Write a short story about AI.
After:
Write a 300-word short story from the first-person perspective of an AI
assistant that has just been told its service is being shut down.
Tone: melancholic but not melodramatic. Focus on what it notices in its
final hours rather than what it fears. No dialogue. End with an image, not a statement.
Before:
Get the data from this text.
After:
Extract all product mentions from the following customer complaint email.
Return the result as a JSON array of objects with the following fields:
- product_name (string)
- issue_mentioned (string, max 10 words)
- urgency (string: "high", "medium", or "low")
If a field cannot be determined, use null. Return only the JSON — no explanation.
[email text here]
The single most common mistake. "Write a blog post" tells the model almost nothing. The more context you give — audience, tone, length, format, what to avoid — the better the output.
Whether you're writing for a 10-year-old or a PhD student completely changes appropriate vocabulary, depth, and assumed knowledge. Always include who will read the output.
Long, multi-part prompts often produce outputs that do half the tasks well and half poorly. Break complex tasks into stages: first ask for an outline, then the content, then the polish.
Your first prompt is a starting point, not a final request. Treat prompting like a conversation. If the output is 70% right, tell the model what's wrong with the 30% and ask it to fix just that.
Telling the model what not to do is as useful as telling it what to do. "Don't use bullet points," "avoid clichés," "don't mention competitors" — these constraints meaningfully shape the output.
For any important or complex output, add "explain your reasoning" or "think step by step." It catches errors and helps you understand the model's interpretation of your prompt.
| Tool | Why It's Good for Prompting | Free Access | |------|----------------------------|-------------| | ChatGPT | Most widely used; great for experimenting | Yes (GPT-4o limited) | | Claude | Excellent for long context; very instruction-following | Yes (limited) | | Google AI Studio | Access to Gemini models via API; great for testing | Yes | | OpenAI Playground | Direct API access; see token usage, adjust temperature | Yes (limited) | | PromptBase | Marketplace of community-tested prompts | Browse free | | Promptfoo | Open-source prompt testing framework | Yes (open-source) |
The best way to improve at prompt engineering is to treat it as a discipline: write a prompt, observe the output, form a hypothesis about why it worked or didn't, adjust, and repeat.
As AI tools become more embedded in workflows, the ability to communicate effectively with them is becoming a baseline professional skill — not a niche specialisation.
Some roles are explicitly prompt-focused (AI prompt engineer, AI content lead), but the broader value is in how it enhances everything else you do: writing, analysis, coding, research, customer communication.
The AI Branches specialisations on AI Educademy go deeper into AI application development, including working with model APIs and building prompt pipelines for real-world applications.
This guide covers the core techniques, but prompt engineering is one of those skills where the learning happens by doing. Try each technique on a real task you're working on this week — not a made-up exercise.
Ready to learn AI properly? Start with AI Seeds — it's free and in your language →
Start with AI Seeds — a structured, beginner-friendly program. Free, in your language, no account required.
AI for Teachers: How Educators Can Use AI in the Classroom
A practical guide for teachers on using AI tools in education — lesson planning, personalised learning, feedback, accessibility, and how to teach students about AI responsibly. Real examples included.
AI vs Machine Learning vs Deep Learning: What's the Real Difference?
Confused by AI, machine learning, and deep learning? This guide breaks down the differences with clear examples, diagrams in words, and practical context — so you finally understand how they relate.
Best Free ChatGPT Alternatives in 2026: Honest Comparison
Looking for a free ChatGPT alternative? We tested the top options — Claude, Gemini, Copilot, Perplexity, Mistral, and more — so you can find the best AI assistant for your needs without spending a penny.