Vague prompts get vague answers. Today you swap fuzzy words for real details and watch your Specificity score climb.
Info
This is Part 3 of Score Your Prompt: The 7-Day Challenge. New here? Start at Part 1. Up next: Structure — Tone, Output Format, and What to Avoid.
Welcome to Day 2. Yesterday you found your baseline. Maybe Completeness felt good after Part 2. Now we hunt the points hiding in plain sight.
Today is about Specificity. It is worth 25 points on the free prompt scorer. It rewards prompts that name real things: a real audience, real numbers, real context.
Here is the truth. Most prompts fail here not because they are short, but because they are blurry. We fix the blur today.
Let's rewrite one prompt and re-score it. You will see the jump for yourself.
Why vague prompts get vague answers
An AI model fills in the gaps you leave. That is its job.
When you write "make this better," the model guesses what "better" means. It might pick shorter. It might pick fancier. You had something in mind, but you never said it.
So you get an answer that is fine for someone, just not for you.
Specificity is how you stop guessing games. You hand the model the exact target instead of a fuzzy direction. The clearer your target, the closer the answer lands.
This is one of the core habits of good prompt engineering. It costs you a few extra words. It saves you three rounds of rewrites.
Step 1: Hunt your vague words
Open the prompt you scored yesterday. Read it slowly. Circle every word that could mean ten different things.
These are the usual suspects:
- "good," "better," "nice," "great"
- "short," "long," "detailed"
- "professional," "casual," "engaging"
- "soon," "a few," "some," "a lot"
- "people," "users," "customers" (without saying which)
Each of these is a door the model walks through alone. Your job is to walk through it with the model.
Tip
A fast test: ask "could two people read this word and picture different things?" If yes, it is vague. Replace it with a number, a name, or an example.
Step 2: Name the exact audience
This is the single biggest specificity win. Tell the model who the answer is for.
"Write about budgeting" could go anywhere. "Write about budgeting for college students living on a part-time paycheck" has a clear reader. The model now picks the right words, examples, and tone on its own.
Audience changes everything downstream. A pitch for engineers reads nothing like a pitch for retirees.
Write a short post about our new app.
Write a 120-word post for busy parents who have never used a budgeting app, explaining how ours tracks spending in under a minute a day.
See the difference? The second one tells the model the reader, the length, the angle, and the benefit. That is four blurry spots made sharp.
Step 3: Swap fuzzy words for numbers
Numbers are the easiest specificity upgrade. They leave no room to guess.
"Short" becomes "under 100 words." "A few tips" becomes "exactly 5 tips." "Make it detailed" becomes "include 3 examples for each point."
Watch how each swap removes a decision the model would otherwise make for you.
| Vague | Specific |
|---|---|
| Keep it short | Keep it under 80 words |
| Give me some ideas | Give me 7 ideas |
| Make it detailed | Add 2 real examples per section |
| Sometime soon | By a 9 a.m. deadline tomorrow |
| For a wide audience | For first-time home buyers in their 30s |
You do not need numbers everywhere. Use them wherever a word is doing a job a number could do better.
Step 4: Add the context only you know
The model cannot see your world. It does not know your product, your boss, your last email, or your goal.
Context is the background that makes the answer fit your real situation. A little goes a long way.
Ask yourself three quick questions:
What is the goal? (What should this output actually achieve?)
What is the situation? (Who, what, where — the facts the model is missing.)
What does success look like? (How will you know the answer is right?)
Answer those in a sentence or two and paste them in. You just handed the model the context that lived only in your head.
Warning
There is a limit. Do not dump three pages of background "just in case." Add the facts that change the answer, and leave out the rest. Too much noise buries the signal the model needs.
Step 5: Show one quick example
Specificity is not only about telling. It is also about showing.
If you want a certain style or format, paste one tiny example of it. "Write product names like 'SwiftPay' and 'BudgetBuddy'" tells the model your taste in five seconds.
One sample is enough to set direction. You do not need a long list to make this work.
This nudges your prompt toward Enhancement too, which we cover in Part 5. For today, treat the example as a way to make your intent crystal clear.
Your Day 2 rewrite challenge
Time to do it. Take one prompt and run it through all five steps.
Here is a weak prompt to practice the pattern on:
Write an email to my team about the project being late.
Now the rewritten version, with audience, numbers, and context added:
Write a short email (under 120 words) to my 6-person design team.
The website redesign will ship one week late because the client
changed the brand colors twice. Goal: keep morale up and set the
new ship date as next Friday. Tone: calm and honest, not panicked.
Open with the new date, then the reason, then one thank-you line.
Same request. Worlds apart. The second one names the reader, the length, the facts, the goal, and the order. There is almost nothing left to guess.
Tip
Once your rewrite works, save it. A tightened prompt is worth keeping in your prompt library so you never rebuild it from scratch.
Re-score and lock in the win
Now the satisfying part. Paste your rewrite into the prompt scorer and watch the Specificity number move.
Read the suggestions it gives back. They often point at one last vague word you missed. Fix it and run it again. Two minutes, real progress.
Write down your new total next to yesterday's baseline. Seeing the jump is what keeps the streak alive.
You attacked Completeness in Part 2 and Specificity today. Those two cover 60 of the 100 points. You have already moved the needle a lot.
Tomorrow we shape the answer itself: tone, output format, and the things to tell the model to avoid. That is Structure, and it is the next 20 points.
Keep going
Next → Day 3: Structure — Tone, Output Format, and What to Avoid
Or see the full Score Your Prompt: The 7-Day Challenge series.
