Understanding AI · Lesson 2 of 7
What is a prompt?
Learn what a prompt is, the different shapes prompts take, and why a few extra words of clarity can completely change the answer you get.
You're learning as a guest — everything is open. Sign in to save progress, earn XP and pick up where you left off.
The simple definition
A prompt is everything you give an AI model to respond to. Most of the time, that's the message you type into the chat box. It can be a question, an instruction, a block of text you paste in, an example you want copied, or even an image or file.
If AI assistants are controlled with words, the prompt is the steering wheel. Every response you get is the model's best attempt to continue from, or answer, what you gave it.
Prompts come in many shapes
There's no single "correct" form. Here are the most common kinds you'll use:
| Type | What it does | Example |
|---|---|---|
| Question | Asks for information or an explanation | "Why does bread go stale?" |
| Instruction | Asks the AI to do a task | "Write a thank-you note to my neighbor for watering my plants." |
| Transformation | Changes something you provide | "Turn these meeting notes into a bulleted action list." |
| Analysis | Asks for an evaluation | "What are the weak points in this cover letter?" |
| Creative | Asks for something new | "Give me ten names for a cozy neighborhood bakery." |
Most real prompts mix these. "Here are my notes (transformation) — summarize them for my manager (instruction) and flag anything unclear (analysis)" is a perfectly normal prompt.
The prompt is more than your last message
In a chat, the model doesn't only see your latest message. It also sees the earlier conversation: your previous messages and its own previous replies. That whole history acts as part of the prompt. This is why you can say "make it shorter" and the assistant knows what "it" is.
Many apps also add hidden instructions behind the scenes (often called a system prompt) that set the assistant's general behavior. You usually can't see these, but they're one reason the same question can get slightly different styles of answer in different apps.
Note: Each new chat usually starts fresh. Unless a tool has a memory feature turned on, the assistant won't remember details from a conversation you had yesterday, so include what it needs to know.
Why wording matters so much
Consider a request you might make at work:
Summarize this report.
Summarize the report below for my team lead, who has two minutes to read it. Use three bullet points: the main finding, the biggest risk, and the decision we need from her. Keep each bullet under 25 words.
Both are valid prompts. But the first one leaves the AI guessing about almost everything: How long? For whom? What matters most? The model fills those gaps with safe, average choices, so you get a generic summary.
The second prompt answers those questions up front. It names the audience (a busy team lead), the purpose (a quick decision), and the format (three specific bullets, each short). The AI no longer has to guess, so the result is far more likely to be usable as-is.
A real-world example
Say you're a teacher who wants a quick classroom activity:
I teach a class of 12-year-olds. Create a 10-minute warm-up activity that helps them practice identifying adjectives. It should work with just a whiteboard, no handouts, and involve the whole class. Give me step-by-step instructions I can follow while teaching.
Notice that this prompt isn't long or technical. It's just specific: the age group, the time limit, the learning goal, the materials available, and the format of the answer. That's the heart of good prompting.
What "prompt engineering" really means
You'll hear the term prompt engineering. It sounds intimidating, but it simply means the skill of writing prompts that reliably get useful results. For most people, that comes down to a handful of habits:
- Say what you want clearly.
- Give the context the AI can't know on its own.
- Describe what a good answer looks like.
- Read the result, then refine.
That last step matters. A prompt isn't a one-shot exam. If the first answer isn't right, you can reply with a follow-up like "Make it more casual" or "Focus only on the second point." Treat prompting as a conversation.
Watch out: A prompt isn't a magic spell. There are no secret words that unlock a hidden "genius mode." Clear, specific, plain language beats clever tricks almost every time.
Quick recap
A prompt is everything the model sees: your message, the conversation so far, and any files or examples you include. Small additions of context, audience, and format can turn a generic answer into one you can actually use.
Common mistakes
Mistake: Believing there are secret magic words that make the AI smarter.
Fix: Focus on clarity and specificity in plain language; that's what reliably improves results.
Mistake: Expecting a new chat to remember what you discussed in a previous one.
Fix: Restate the key details or paste the relevant information at the start of a new conversation.
Mistake: Giving up after the first answer misses the mark.
Fix: Reply with a specific follow-up ('shorter', 'more formal', 'focus on X') to steer the answer.
Practice exercise
Upgrade a one-line prompt
The starter prompt below is a real request someone might type, but it leaves the AI guessing.
Rewrite it so it answers these three questions:
- Who is the result for?
- Why do you need it (what will you do with it)?
- What shape should the answer take (length, format, tone)?
Keep the rewrite to three or four sentences.
Your draft stays in this browser tab — it isn't sent anywhere.Try your draft in the Playground
Hints
You can do the whole exercise as a guest.Sign in to save progress
Knowledge check