Prompt Engineering Basics (Five Habits That Work)

"Prompt engineering" sounds technical, but it is really just learning to ask well. There is no secret syntax and no magic words — the people who get great results from AI simply give it clearer instructions than everyone else. Five habits cover almost all of it. If you are brand new to the tools, start with ChatGPT for beginners first, then come back here.

1. 01

   State the task and the format you want

   Lead with exactly what you want and what shape it should take. "Summarise this in five bullet points," "Write three subject-line options," "Turn these notes into a short client email." A vague ask gets a vague answer; naming the output — a list, a table, a paragraph, a number of options — removes most of the guesswork before the model starts.

2. 02

   Give it a role and the context

   Tell it who it is being for this task — "you are a careful proofreader," "act as a sceptical reviewer" — and paste in the material it needs to work from. The role sets the tone and the standard; the context is the raw material. Both are things the model cannot infer, and supplying them is what separates a generic answer from one that fits your situation.

3. 03

   Show it one good example

   If you have a version you like — a past email in your voice, a report in your format — paste it in and say "match this style." One good example teaches the model more than a paragraph of instructions, because it shows rather than tells. This is the single most underused habit, and it is the one that gets results that sound like you instead of like a machine.

4. 04

   Constrain length, tone and what to avoid

   Left unbounded, models pad. Set limits: "under 120 words," "plain English, no jargon," "do not use the word ‘leverage’," "Australian spelling." Telling it what to avoid is as useful as telling it what to do. Constraints are not fussy — they are how you get something you can use as-is instead of something you have to trim and fix.

5. 05

   Iterate, then save the winners

   The first answer is a draft. Push back — "tighter," "warmer," "cut the intro," "that is too formal." When one finally lands, save it. The prompts that work for your recurring tasks become a small personal library, and reusing a proven prompt beats writing a fresh one every time. That library, shared across a team, is most of what "getting good at AI" actually is.

People also ask

Do I need to learn prompt engineering as a proper skill?

Not as a technical discipline. For everyday work, the five habits above cover the vast majority of what matters — be specific, give context and an example, set constraints, and iterate. The deeper, more technical side matters mostly when you are building AI into a production system, not when you are drafting an email.

Why do I get generic answers from AI?

Almost always because the prompt was generic. The model fills gaps with the most average, safe response it can, so a vague ask produces a vague result. Add the specifics — audience, purpose, length, tone, an example — and the answer stops being generic because it now has something to work with.