BeginnerLesson 1 of 306 min read

What prompt engineering is and why it matters

See why two people get very different answers from the same AI, learn what a prompt is, how the model reads it, and run your first experiment.

In this lesson you will

  • Explain what a prompt and prompt engineering are
  • Understand how an AI uses your words as clues
  • Compare a vague and a precise prompt
  • Run your first experiment

About 14 minutes

Two people, one AI, very different results

Ayesha and Hamza both need a short announcement for their school’s science fair. Ayesha types: write an announcement. Hamza types a few extra lines. Ayesha gets four generic paragraphs about “an exciting event”. Hamza gets a short, friendly notice with the date, the place, the prize categories and a call to register by Friday. They used the same AI on the same day. The only difference was the prompt. Learning to write the second kind is what this course is about.

What a prompt is

A prompt is whatever you give an AI model so that it can produce a response: a question, an instruction, a pasted document, an example, or all of these together. Prompt engineering is the practice of designing prompts carefully so that the response is useful, accurate and in the shape you need, and doing it in a way you can repeat.

The word “engineering” is a little grand. You do not need a degree or a special language. You need clear thinking, a few reliable patterns and the habit of testing what you write. Those are skills anyone can learn in a few weeks of practice.

How the AI reads your prompt

A chat model is trained on a huge amount of text to predict what text should come next. When you write a prompt, the model uses every word of it as clues about what kind of answer you want. It cannot see into your head, your school, your job or your plans. It only has the words you typed (and, in some tools, earlier messages and settings). Three consequences follow:

  • Missing details get filled with guesses. If you do not say who the announcement is for, the model invents an audience.
  • Vague words give vague answers. “Good”, “nice” and “professional” mean different things to different people.
  • The model is eager to please, not to question. Most tools will produce something even from a poor prompt. A confident answer is not proof that it understood you.

The same task, two prompts

Write an announcement.
Exciting news! We are thrilled to announce an upcoming event that you will not want to miss. Join us for a day full of fun, learning and discovery. Stay tuned for more details!
Write a 60-word announcement for parents about our school science fair on Friday 14 March in the main hall, 10 am to 1 pm. Mention three prize categories (Best Experiment, Best Poster, Most Creative). Friendly tone. End with: register your child by Wednesday at the front office.
Parents, please join us for our School Science Fair on Friday 14 March, 10 am to 1 pm, in the main hall! Students will show their experiments and posters, and prizes will go to Best Experiment, Best Poster and Most Creative. Come and cheer them on. Please register your child by Wednesday at the front office.
Good to know

The replies above are written examples to show the idea. Your AI will word things differently, but the pattern holds: specific prompts get specific answers.

What prompt engineering is not

  • It is not magic words. There are no secret phrases that unlock hidden power. Clear information beats tricks.
  • It is not only for programmers. Teachers, marketers, students and managers all benefit.
  • It is not a replacement for knowing your subject. You still judge the result.
  • It is not one perfect prompt. It is a loop: write, test, improve.

What you will be able to do after this course

Level You will be able to
Beginner (lessons 1 to 10) Write clear prompts with role, task, context and format; use examples; improve a prompt step by step; build your own starter pack
Intermediate Break big tasks into steps, ask for structured output, reduce errors, and build reusable templates
Advanced Test and compare prompts, handle tools and long documents, and design prompts for other people’s products
Projects and go pro Build a portfolio and use prompt skills in a job or freelance work

Tools you need

Any AI chat tool works for this course: the free versions of ChatGPT, Claude, Gemini, Copilot or similar. The ideas apply to all of them. Features, limits and names change often, so check each tool’s own help pages for details. Do not paste private information such as passwords, ID numbers, medical details or other people’s personal data into any AI tool.

Your first experiment (10 minutes)

  1. Pick a small task you really have: a message to send, a list to make, a topic to understand.
  2. Write the laziest prompt you can, and save the answer.
  3. Now write a prompt that adds: who it is for, what it must include, how long it should be, and the tone.
  4. Compare the two answers. Which one could you use as it is? Write down what changed.

A tiny checker you can run

To make the idea concrete, here is a short program that looks for four useful ingredients in a prompt. It is only a rough word check, not a real judge of quality, but it shows how much more a good prompt carries. The output is real:

def ingredients(prompt):
    p = prompt.lower()
    checks = {
        "audience": any(w in p for w in ("for parents", "for students", "for customers", "audience", "for beginners", "for my")),
        "length": any(ch.isdigit() for ch in p) and "word" in p,
        "tone": any(w in p for w in ("friendly", "formal", "casual", "polite", "tone")),
        "details": sum(ch.isdigit() for ch in p) >= 2,
    }
    return checks

vague = "Write an announcement."
precise = ("Write a 60-word announcement for parents about our school science fair on "
           "Friday 14 March, 10 am to 1 pm. Friendly tone.")
for name, text in (("Vague", vague), ("Precise", precise)):
    found = ingredients(text)
    print(name, "->", sum(found.values()), "of 4:", [k for k, v in found.items() if v])
Vague -> 0 of 4: []
Precise -> 4 of 4: ['audience', 'length', 'tone', 'details']
Common mistake

Judging a prompt by how clever it sounds. Judge it by whether the answer is usable. A short, plain prompt that gets the right result is better than a long, fancy one.

Keep an experiment notebook

Save your prompts and the answers in one note or document, with a line on what worked. Within a week you will have your own proven patterns to reuse.

Check what you learned

Five quick questions. Pick an answer and you will see straight away whether it is right and why.

What is a prompt?

  1. Only a question mark
  2. Whatever you give an AI so it can respond
  3. A kind of password
  4. A paid plan

Questions, instructions, examples and pasted text are all prompts.

Why do vague prompts give vague answers?

  1. The model refuses to work
  2. Vague words are banned
  3. The internet is slow
  4. The model fills gaps with guesses

It only has the words you typed.

Which statement about prompt engineering is true?

  1. It is a loop of write, test and improve
  2. It needs secret magic words
  3. It is only for programmers
  4. One perfect prompt works forever

Testing and improving is the core habit.

What should you never paste into an AI tool?

  1. A public recipe
  2. Your own essay topic
  3. Passwords and private personal data
  4. A made-up example

Keep private information out.

A confident answer from an AI proves that…

  1. It understood perfectly
  2. It is always correct
  3. It is from the internet
  4. Nothing by itself; you still check it

Confidence is not evidence.

Summary

  • A prompt is everything you give the AI; prompt engineering is writing it carefully and testing it.
  • The model only has your words, so missing details become guesses.
  • Specific prompts get specific answers.
  • Write, test, improve, and keep a notebook.

Next lesson

Next: the anatomy of a prompt, the four parts every strong prompt can use.

Questions

Do I need to know programming?

No. Prompt engineering is clear writing plus testing. Programming helps later for building automations, but it is not required.

Which AI tool should I use?

Any major chat assistant works. Start with the free version of one you can access and try others later.

Is prompt engineering a real job?

Some roles exist, but the skill is mostly valuable as part of other jobs such as marketing, writing, support, teaching and development.

Progress is saved in this browser.