Most people prompt AI like they’re making a wish. One vague line, a hopeful click, then disappointment when the answer comes back generic. The better approach treats the prompt like a short brief for a smart new hire, and the results get noticeably sharper. Most people hand over a sentence and hope. This approach hands over a brief, and it’s faster in the end because you skip three rounds of “no, not like that.”
A quick note before we start. The Reddit post this came from was just a link to a beginner video on prompting basics, with no discussion yet. So I’m not summarizing the video. This is the practical version of the same idea, so you can use it today.
The key idea
AI doesn’t read your mind. It reads your words. Whatever you leave out, it fills in with the most average guess. Better prompts aren’t clever. They’re just more complete.
Think about your first day at a new job. If your manager said “handle the newsletter,” you’d have a dozen questions. Who reads it? How long should it be? What happened last time? A smart new hire would ask. The AI won’t ask. It will just start writing and hope for the best. That’s why the answer feels generic: you gave it nothing specific to work with, so it gave you nothing specific back.
Old way vs new way
The old way: “Write me a marketing email.”
The new way: “You’re a copywriter for a small coffee subscription company. Write a 120-word email to people who abandoned their cart. Friendly tone, one clear call to action, no discounts over 10%.”
The first prompt makes the AI guess your industry, audience, length, tone and goal. The second one removes every guess. Same tool, completely different output, and you spent maybe 30 extra seconds.
Here’s a second example outside of marketing. Old way: “Help me prepare for my interview.” New way: “You’re a hiring manager at a mid-size logistics company. I’m interviewing for an operations analyst role. Ask me five tough questions one at a time, wait for my answer, then tell me what was weak about it.” The first version gets you a generic list of tips you’ve read a hundred times. The second one gives you a live practice session.
The old way also treats the first answer as final. The new way treats it as a draft. You react to it, tighten it, and ask again. Most of the quality comes from that second and third pass, not from the first message.
Practical steps 🛠️
- Give it a role. “You’re a senior recruiter” or “You’re a patient math tutor” sets the voice and the level of detail. Pick a role that matches the kind of expert you’d actually hire for the job.
- State the task in one sentence. What do you actually want produced? Be specific about the format: email, table, list, script. If you can’t say it in one sentence, the task is probably two tasks, so split them.
- Add context. Who is this for? What do they already know? What’s the goal? Two or three sentences here usually beat a long paragraph of background.
- Set constraints. Word count, tone, things to avoid. Constraints are where quality comes from. A useful trick is to list what you don’t want, such as “no buzzwords” or “no intro paragraph,” because bad habits are easier to block than to describe.
- Show an example. Paste one sample of what good looks like. This is often the single biggest upgrade. If you don’t have one, paste something close and say “similar length and tone, different topic.”
- Iterate. Say what to change: “shorter,” “less formal,” “cut the first paragraph.” Small nudges beat rewriting everything. Keep the conversation going in the same thread so the AI remembers what you already agreed on.
You don’t need all six every time. For a quick question, a role and a constraint might be enough. For something you’ll send to a customer or a boss, use the full list.
A fast test
Before you hit enter, ask yourself: could a stranger read this prompt and know exactly what I want? If a human would need to ask follow-up questions, the AI needs those answers too.
Try it on a coworker if you can. Hand them the prompt on paper and ask what they’d produce. If their version looks nothing like what you pictured, the prompt has gaps. Fix those gaps first, and you’ll rarely need to fix the output.
Your move
Take one prompt you used this week that gave a mediocre answer. Rewrite it using the six steps above and compare the two outputs side by side. You’ll see the difference right away. If you want more guides like this, stick around, because there’s plenty more where this came from!