There is no shortage of prompt engineering advice. There are YouTube channels, Substack newsletters, courses with 47 modules, and a thousand LinkedIn posts promising "the 10 ChatGPT prompts that will change everything." Most of it is noise.
Not because the tips are wrong, exactly. Some of them work. But they share a problem: they give you fish instead of teaching you to fish. Memorising 20 specific prompts doesn't help you when you face a situation none of those prompts fits. What helps you is understanding why prompts work, at a structural level, so you can construct a good one for any situation.
Here's the honest reduction. After all the tutorials, all the experiments, and all the hours people have spent refining this: there are three prompt structures that cover the vast majority of useful things a beginner wants to do with AI. You don't need the other 44 tutorials. You need these three, practiced until they're automatic.
This module teaches them with examples from personal and career contexts, since that's what most people actually need at this stage.
Pattern 1: Role + Task + Constraints + Format
This is the most broadly useful structure in the language. When in doubt, use this one.
What it is: You tell the AI what role to play, what you need it to do, what constraints to work within, and what format you want the output in.
Why it works: Most bad prompts fail on exactly one of these four dimensions. Either there's no clear task ("help me with my career"), no constraints ("write something professional"), no format guidance (you get an essay when you wanted three bullet points), or no role context (the AI defaults to a generic helpful assistant mode when a more specific framing would serve you better).
The structure:
You are [role, a specific type of expert or perspective].
Your task is to [clear, specific task].
Constraints: [what to include, exclude, aim for, avoid].
Format: [exactly how you want the output structured].Bad version:
Can you help me prepare for a job interview?This gets you a generic list of interview tips that reads like a careers website from 2015.
Good version:
You are a senior hiring manager who has interviewed 200+ candidates for
mid-career roles in project management.
Your task is to help me prepare for a panel interview for a Project
Coordinator role at a mid-sized construction company in Melbourne.
Here's my background: I'm currently an office manager with 6 years of
experience. I've coordinated projects informally but have no formal PM
qualifications.
Constraints: Focus on the questions I'm most likely to struggle with,
not the easy ones. Be honest about the gaps in my background and how
to address them without bluffing.
Format: Give me 5 likely hard questions, and for each one: the subtext
behind the question (what they're really asking), and a 2 to 3 sentence
suggested answer framework I can adapt.The second version tells the AI who it's talking to, what the actual situation is, and precisely what format will be useful. The output is unrecognisable from the first version.
Try it for: Interview prep, writing a difficult email, explaining a complex topic, planning a hard conversation, structuring a pitch.
Pattern 2: Examples-first
What it is: Instead of describing what you want, you show the AI an example of the output you're after, then ask it to produce something in that style.
Why it works: Writing style, tone, and format are extraordinarily hard to describe in words. "Professional but warm" means something different to everyone. "Like how I normally write" is even harder to convey abstractly. But if you paste in an example, the AI can pattern-match to it with surprising accuracy.
The structure:
Here's an example of the kind of [thing] I want:
[paste the example]
Now write [your actual thing], in the same style, tone, and structure as
the example above. The content should be: [your specific brief].Bad version:
Write a LinkedIn post about my experience moving from nursing into
healthcare administration. Keep it professional but personal.This produces a serviceable but generic post. You've described the vibe, but "professional but personal" is doing a lot of heavy lifting.
Good version:
Here's an example of the kind of LinkedIn post I'm going for:
"Five years ago I was asking patients if they needed more blankets.
Last week I was presenting a capacity report to the hospital board.
I didn't plan this career. I stumbled into it by being the nurse who
couldn't stop asking 'why does this process exist?' Eventually someone
said, 'You should be in management.' I didn't believe them for about
a year.
If you're in clinical work and feel a pull toward the operational side,
it's not betrayal. It's a different way to care for the same people."
Now write a LinkedIn post in this same style and tone about my
experience transitioning from [your transition]. Key details I want
to include: [specific points from your actual story].The example doesn't even have to be your own writing. Find a post you've seen and genuinely liked, paste it in, and use it as the stylistic template.
One constraint: Don't paste in copyrighted content that you're going to claim as original. Use the example as a style reference, not something to copy.
Try it for: LinkedIn posts, About sections, cover letters, emails you need to write in a specific tone, social media captions.
Pattern 3: Critique-then-rewrite
What it is: Instead of asking AI to produce something from scratch, you show it what you've already written (or a first draft), ask it to critique it honestly, then ask it to rewrite with those critiques in mind.
Why it works: Two reasons. First, your existing draft already contains information about you, your voice, and your actual content, which gives the AI much more to work with than a blank brief. Second, seeing the critique first means you can agree or disagree with each point before the rewrite happens. If the AI says "your third paragraph is too cautious," you can decide if that's right, and if it is, you've also learnt something about your own writing.
The structure:
Here's something I've written. I want you to do two things.
First: critique it. Be honest and specific. Tell me what's working,
what isn't, and what a sharp editor would cut or rewrite. Don't be
gentle, I'd rather hear the problems now than later.
Second: rewrite it, applying your own critique. Keep my voice and my
core content. Don't add things I didn't say. Just make it better.
Here's the text:
[paste your draft]Bad version:
Can you improve my cover letter? [pasted text]"Improve" is vague. The AI will polish it. It might not actually fix the structural problem, which could be that you buried the lead, or that the whole thing sounds like you're apologising for applying.
Good version:
Use the critique-then-rewrite structure above. You'll get a specific list of what's not working, then a rewritten version that addresses those exact points. And you'll be able to evaluate whether each change is actually better, which builds your own judgement over time.
Try it for: Cover letters (do this one, it makes a real difference), LinkedIn bios, study notes you want to compress into a clearer summary, emails you've written that feel off but you can't work out why, any writing you want to improve without losing your voice.
What doesn't work
Combining all three patterns into one mega-prompt on your first attempt. They work best used one at a time. Master each one before layering.
Using Pattern 3 on something you're not willing to change. If you're going to read the critique and defend every sentence anyway, you're wasting both your time and the AI's. Go in willing to hear something uncomfortable.
Using Pattern 1 with a role that's too vague. "You are an expert" is nearly useless. "You are a senior HR manager at a professional services firm who has reviewed 500+ graduate applications" gives the AI somewhere real to stand.