How to write prompts that get you what you want

To write AI prompts that get you what you want, give the model three things: context (who you are, what this is for), one example of what good looks like, and permission to ask you questions before answering. That combination fixes most bad AI output on the spot. Everything else in this post is me showing you how those three moves work in real life, because I use them every single day and I built an entire app on the back of them.

No engineering degree required. I checked.

Stop talking to it like a search bar

Most people type into ChatGPT or Claude the way they type into Google: three words and a prayer. “Marketing plan wellness app.” Then they get something generic and decide AI is overhyped.

The model isn’t a search bar. It’s closer to a very fast, very eager new hire who started this morning and knows nothing about you. Would you tell a new hire “marketing plan wellness app” and walk away? You’d brief them. You’d say who the audience is, what’s been tried, what you hate, when it’s due.

That briefing is the prompt. The quality of what comes back tracks the quality of what you put in, almost embarrassingly closely.

The three ingredients, in order

Context. Tell it who you are and what the output is for. “I’m the founder of a gut health app for women. I’m writing a welcome email for people who just downloaded it. They’re skeptical of wellness marketing.” Two sentences of context routinely doubles the quality of the answer. I have watched this happen in workshops so many times it’s almost a party trick.

An example. Models are imitators before they are inventors. Paste in one email you love, one report formatted the way your boss likes, one caption that sounds like you. Then say “match this tone and structure.” Describing your style in adjectives (“friendly but professional”) gets you beige. Showing it gets you something recognizable.

Permission to ask questions. This is the one nobody uses and everybody should. End your prompt with “ask me up to five questions before you start.” The model will surface exactly the gaps you didn’t know you left. Half the time, answering its questions clarifies the task for you too.

Iterate like it’s a conversation, because it is

When I was building Gut Goddess, my food tracking app, there was a screen I made the AI redo eleven times. Eleven. Not because the AI was stupid, but because my first prompt was. I said “make it cleaner” when what I meant was “less text, bigger buttons, and stop making everything teal.”

The lesson stuck: when the output is bad, read your own prompt back before blaming the model. Then don’t start over. Reply. “Shorter.” “Less formal.” “You lost the point about pricing, put it back.” Each correction teaches it what you meant, and by round three you’re usually there.

Once you land on a prompt that works, save it. A prompt you can reuse every Monday is worth more than a clever one you wrote once and lost.

What to skip

You do not need “prompt engineering” courses, magic phrases, or those 400-word template monsters people sell on the internet. You don’t need to say please, although I do, because I’m apparently raising the AI to have manners.

You need context, an example, questions, iteration. That’s the whole method. The rest is practice on your own real work, which is exactly where most people never take it.

FAQ

What makes a good AI prompt?

Context about you and the task, a concrete example of the output you want, and an invitation for the AI to ask clarifying questions first. Specific beats clever every time.

Why are my AI results so generic?

Because the prompt was generic. If you give the model no context and no example, it defaults to the average of the internet, and the average of the internet is beige. Brief it like a new hire and the output changes immediately.

Do I need to learn prompt engineering to use AI well?

No. The fundamentals in this post cover most real work. Formal prompt engineering matters if you’re building AI products, not if you’re writing emails, plans, and reports with it.

I teach this live, on your actual work, in 1-1 AI sessions and team workshops. You bring the task you’re stuck on, we build the prompts together, you leave with ones you’ll reuse for months. If that sounds better than reading another listicle of “50 magic prompts”.. you know where my contact page is. Go say hi.