The single biggest upgrade
Here is the secret that takes most people from "meh" results to genuinely useful ones: be specific. An AI model like ChatGPT, Claude, or Gemini cannot read your mind. When you type something short and fuzzy, the model fills in the gaps with the most average, middle-of-the-road guess it can make. That is why a vague prompt almost always produces a bland, forgettable answer.
Think of it like ordering at a coffee shop. If you walk up and say "a coffee, please," you will get whatever the barista decides is normal. But if you say "a large oat-milk flat white, extra hot, one sugar," you get exactly what you wanted. The model is the same: the more clearly you describe the cup you want, the closer it pours.
There are three details that do most of the heavy lifting. Get these three right and your prompts instantly improve:
- WHAT you want — the actual task and any specifics. "Write a 5-bullet summary," not just "summarize."
- WHO it is for — the audience or reader. A summary for your boss reads very differently from one for a 10-year-old.
- In WHAT form — the format you want back: a table, an email, a checklist, three options, 100 words.
The model gives you the average of all possible answers unless you narrow it down. Specifics are how you steer.
Vague vs clear, side by side
Let's make this concrete. Imagine you want help writing a message to your team about a meeting moving to next week. Here is the version most people type first:
Write something about the meeting changing.The model has no idea who is reading it, how formal it should be, or how long. You will get a generic paragraph that you then have to rewrite anyway. Now watch what happens when you add the three details:
Write a short, friendly email to my 6-person design team letting
them know our Monday 10am standup is moving to next Tuesday at 2pm.
Keep it under 80 words, warm but professional, and end with a
one-line call to confirm they can make it.See the difference? The second prompt names the what (the meeting change and the new time), the who (a 6-person design team), and the form (a short, warm email under 80 words ending with a confirmation line). You will get back something you can almost send as-is. The extra fifteen seconds of typing saves you several minutes of editing.
If you find yourself heavily rewriting the answer, that is a signal your prompt was missing a detail. Add it and run again, rather than fixing the output by hand.
A one-line formula you can use today
You do not need to memorize a long checklist. Here is a tiny formula that bakes in everything above. Reach for it whenever you are stuck:
[Role] + [Task] + [Context] + [Format]
- Role — who you want the AI to act as. "You are a friendly fitness coach." This sets the tone and expertise.
- Task — the one clear thing to do. "Create a 3-day beginner workout plan."
- Context — the facts the model needs about your situation. "I'm 40, work from home, and have only dumbbells."
- Format — how the answer should be shaped. "Show it as a table with day, exercise, and reps."
Put together, the formula produces a prompt like this — copy it, swap in your own details, and you are off:
You are a friendly fitness coach for total beginners. [Role]
Create a 3-day-per-week workout plan I can do at home. [Task]
I'm 40, work at a desk all day, and only own a pair of
dumbbells and a yoga mat. [Context]
Give it to me as a table with columns for Day, Exercise,
Sets, and Reps, plus one short tip at the end. [Format]You don't always need all four blocks. For a quick factual question, Task alone is fine. The formula is a safety net for when answers come back weak — not a rule you must fill out every time.
This works across every major model. ChatGPT, Claude, and Gemini are all trained to follow clear instructions, so the same well-structured prompt tends to give strong results no matter which one you open.
Time to feel the difference yourself. This takes about five minutes and gives you the first entry in your personal prompt collection.
- Pick one task from Lesson 1.1 — something you'd genuinely use AI for, like drafting an email, planning a trip, or summarizing notes.
- Write a vague version: just a few words, the way you'd normally rush it. Run it in any model (ChatGPT, Claude, or Gemini) and skim the result.
- Now write a clear version using the formula: [Role] + [Task] + [Context] + [Format]. Spell out the what, who, and form.
- Run the clear version in the same model and compare the two answers side by side. Notice how much less editing the second one needs.
- Copy the better prompt into a notes file or doc titled "My Prompt Collection." This is the start of a library you'll keep reusing.
📌 Save your winning prompt to your prompt-doc now — future-you will thank you when you need it again.
- Vague prompts get average answers; specifics are how you steer the model toward what you actually want.
- The biggest upgrade is naming what you want, who it's for, and in what form.
- Use the formula [Role] + [Task] + [Context] + [Format] whenever an answer comes back weak.
- Save your best prompts — a reusable collection turns one good result into many.