How do I write a prompt that gets a good answer?

Write it like a brief, not a search query. Say what you want, who it's for, what to avoid, and what good looks like — and show an example rather than describing one. A vague answer is almost always the result of a vague ask.

Do I actually need to learn "prompt engineering"?

Not as a discipline with a certificate. What you need is the habit of briefing properly — saying who it's for, what it's for, what shape you want it in, and what good looks like. That's four sentences, not a course. The genuinely technical end of the field has also moved on: for anyone doing serious work with these tools, the leverage is now in curating what the model can see rather than in wording the request cleverly.

Are there magic phrases that make the output better?

Very few, and they age badly. Tricks like offering tips or insisting the model take a deep breath came from a specific generation of models and mostly don't survive contact with current ones. The things that reliably work aren't tricks at all, because they're just information transfer: giving context, giving examples, giving the actual source material, and asking for a plan before the work. Those keep working precisely because they're not exploits.

Should my prompt be long or short?

As long as the useful information, and no longer. Length itself doesn't help — a long prompt padded with restatements and pleading is worse than three specific sentences. But under-specifying is by far the more common failure: most disappointing answers come from a one-line request that left the model to guess the audience, the format, and the standard. Specific beats short; specific and short beats everything.

What do I do when it keeps ignoring an instruction?

Assume it's buried rather than defied. Instructions in the middle of a long block of text carry much less weight than ones at the start or end, and a single line in a wall of context is easy to lose. Move the requirement to its own line at the end, state it as a hard constraint rather than a preference, and — if it's genuinely non-negotiable — ask the model to confirm the constraints back to you before it starts.

Asking for what you actually want

There is no secret syntax, no magic words, and nothing to memorise. A prompt is just what you type; prompt engineering is the unglamorous craft of typing it clearly. If that sounds like it can’t possibly account for the difference between a useless answer and an excellent one, it does — and that gap is the cheapest thing you will ever fix about your AI results.

The reframe that does most of the work: you are not searching, you are briefing. Search rewards short keywords because a search engine is matching, not producing. Here you’re handing a capable, fast, slightly over-eager contractor a piece of work — one who has never met your company, doesn’t know who the audience is, and will not stop to ask you a clarifying question unless you invite one. Everything you’d tell that person, you should type.

Two habits raise quality more than any wording trick. Show instead of describing: pasting one paragraph in the tone you want beats three adjectives about tone, every time. And ask for the plan before the work on anything longer than a few minutes — reading a five-line outline and correcting it costs you nothing, while reading a finished draft built on the wrong assumption costs you the whole draft.

The useful thing to know is where this stops being the lever. If the answer is badly written, that’s a prompting problem. If the answer is uninformed — it didn’t know your pricing, your customers, your file — no amount of rewording will fix it, because the material was never in the room. That is a context problem. And when you find yourself typing the same standing instructions into every session, they belong in a project memory file instead of your fingers.

The words

Prompt: A prompt is simply what you say to an AI — the question, instruction, or request you type in. "Rename these photos by date" is a prompt. There's no secret syntax; clear plain English is the whole skill.
Prompt engineering Prompting: The craft of writing instructions that actually get the result you want from an AI — being specific instead of vague, giving examples instead of just describing, and showing the format you expect rather than hoping it guesses. It's less about secret tricks and more about clear communication: say what you want, what to avoid, and what 'good' looks like.

How to ask AI for what you actually want You're not typing a search query. You're briefing a fast, capable, slightly over-eager contractor who has never met your company and will never ask you a clarifying question unless you invite one. 12 min

How to do it

What you want	On Desktop (no code)	Command line (Power Track)
Ask Claude to do something	Type it in the chat, in plain words	`claude "organize this folder"`

The words

Read more

How to do it