Most people type a question into an AI tool the same way they'd type it into Google. Short, vague, a couple of words. They get a mediocre answer, shrug, and think "I guess AI isn't that impressive."
The problem isn't the AI. The problem is the prompt.
A prompt is simply everything you type to the AI before it responds. But what you put in that box — how much detail, in what order, framed in what way — determines everything about what comes back.
"A bad prompt is like calling a contractor and saying 'build me something.' A good prompt is like handing them blueprints, a budget, a deadline, and your favourite materials."
This isn't a small difference. We're talking about the difference between a useless paragraph and something you'd be happy to send straight to a client. Between a confusing explanation and one that clicks immediately. Between wasting 20 minutes and saving two hours.
Let's look at real examples. Select a scenario, then toggle between the bad and good version to see the difference.
You've seen the gap. Now let's name what created it. Good prompts consistently do four things that bad prompts don't:
Good prompts explain the situation. Who are you? What's the background? Why does this matter? Context is the raw material the AI works with.
Not "help me with X" but "write / analyse / summarise / compare / create." A precise verb changes everything. The AI knows exactly what action to take.
Should it be a bulleted list? A 200-word paragraph? A table? Three options to choose from? Without format guidance, the AI guesses — and often guesses wrong for your needs.
Is this for a technical expert or a 10-year-old? Is it formal or casual? Empathetic or direct? Telling the AI who will read this changes how it writes.
After looking at thousands of prompts, the same mistakes come up again and again. Recognising these in your own writing is the first step to fixing them.
"Write me a marketing email." Full stop. No product, no audience, no goal, no tone. The AI has to invent everything — and it will, just not in the way you wanted.
"Make this better." Better how? More concise? More persuasive? Better grammar? More emotional? Vague requests get vague improvements.
"Write a 500-word blog post that is SEO-optimised, funny, professional, targeted at beginners, experts, and includes 5 examples but is also concise." Contradictory instructions produce confused outputs.
Great prompting is a conversation, not a single request. If the first response isn't what you wanted, that's not failure — that's iteration. More on this in Lesson 3.
Stop thinking of prompts as search queries. Think of them as briefs you'd give to a very capable assistant. The more clearly you brief them, the better the output. You're not just asking a question — you're directing a collaboration.
The AI is capable. The prompt is the lever. Every improvement you make to your prompts multiplies the quality of everything you get back.
Think about the last time you used an AI tool and felt underwhelmed by the response. What did you actually type? Now, using the four principles above, rewrite that prompt with more context, a precise task, a format, and an audience.
Open Claude and try both versions back-to-back. The difference will be immediately obvious.
Try it on Claude →