The Five Ingredients of a Good Prompt

A prompt is just what you type into the chat. The word sounds technical, but it is not. Anything you type is a prompt — a question, a request, a half-formed thought. The skill of prompting is simply writing prompts that get you what you actually want, the first time or in just a few rounds.

Here is the whole game in one sentence: a vague prompt gets a vague, generic answer, and a rich prompt gets one built around you. Same AI, same topic, wildly different results — and the only thing that changed is what you put in.

Why this works — you're steering a prediction

Here's the part almost no course explains, and it's the thing that makes everything else obvious. An AI doesn't look your answer up. It predicts, word by word, the most likely continuation of the text you gave it. Your prompt is the beginning of that text. So a vague prompt like 'give me tips on saving money' could plausibly continue ten thousand different ways — advice for a student, a retiree, a millionaire — and with nothing to distinguish them, the model lands on the blandest average of all of them. That's not the AI being lazy. That's the mathematically safest guess when you've told it almost nothing.

Every concrete detail you add quietly rules out thousands of those generic continuations and steers the prediction toward the answer you actually want. You're not commanding the AI, and you're not tricking it — you're narrowing the field of likely responses until the one you need becomes the obvious next thing to say. Hold onto that one idea: good prompting is narrowing the possibilities. It's the lens behind all five ingredients, and behind every prompt you'll ever write.

There is a small industry selling courses on 'prompt engineering.' Most of it is noise — lists of magic words to memorize. But once you understand that you're steering a prediction, you don't need the lists. The five ingredients below are simply the five most powerful ways to narrow the field, and you can apply them to anything — an email, a recipe, a lesson plan, a tough decision.

Let's take the five ingredients one at a time. For each, notice not just what it is, but what goes wrong when you leave it out — because that's what you'll feel in your own results.

1. Context — who you are and the situation you're in

The AI cannot read your mind. A question that's obvious to you is a mystery to a stranger who just walked into the room. Two sentences about your situation — your job, your goal, what you've already tried — can change the answer completely. Leave it out and you get advice written for 'the average person,' which is to say, for no one.

2. Task — the specific thing you want done

Be explicit about the deliverable. Should it explain, compare, draft, summarize, quiz you, make a plan? Lead with a verb: 'Write…', 'Compare…', 'Give me a plan for…'. 'Tell me about saving money' is a topic, not a task — and a topic gets you a lecture, not help.

3. Format — how you want the answer laid out

A numbered list. A short paragraph. A table. A step-by-step checklist. AI will give you almost any shape you ask for — but if you don't say, it defaults to a medium-length essay you then have to reshape yourself. Naming the format is the difference between something you can use immediately and something you have to clean up.

4. Audience — who it's for, and at what level

'Explain it like I'm a smart twelve-year-old.' 'Write this for a busy executive who values directness.' 'I'm brand new to this — no jargon.' Telling the AI who the answer is for is responsible for more 'that's exactly what I needed' moments than any other single move.

5. Constraints — what to limit, keep, or avoid

'Under 200 words.' 'No clichés.' 'Don't suggest anything that costs money.' Constraints feel like they'd box the AI in, but they do the opposite — they free it to focus on what you actually care about instead of trying to cover everything.

Now watch what the ingredients actually do. Here is the same topic — saving money — asked two ways.

It works for anything, not just money

The five ingredients aren't a money trick — they're how you get a useful answer out of AI for almost anything. Here's the exact same recipe aimed at a completely different job: a tricky work message.

A few traps to avoid

• Cramming everything into one dense run-on. Keep each ingredient to a short phrase — the AI reads structure just fine.
• Contradicting yourself: 'keep it brief' and 'be thorough' in the same prompt. Pick one, and say which matters more.
• Being vague about format ('make it nice'). Name the real shape: three bullets, a table, a short paragraph.
• Skipping Audience. 'Explain compound interest' and 'explain compound interest to a curious 10-year-old' are worlds apart.

Recap

• Context — who you are and the situation you're in.
• Task — the specific thing you want, led by a verb.
• Format — the shape you want the answer in.
• Audience — who it's for, and at what level.
• Constraints — what to limit, keep, or avoid.
• You won't always need all five — but Context and Audience are the two that make an answer truly yours.

Here is a weak, vague prompt: 'Give me some tips on saving money.' Rewrite it into a strong prompt that uses all five ingredients — and make it about your real situation, not a made-up one.

Run your rewritten prompt in a chatbot, then bring back both your prompt and the answer it gave you.