Write Prompts That Get You Better Answers
Learn a repeatable way to brief AI clearly, judge its first answer honestly, and improve it on purpose — proven on a real message you actually need to write.
What You're Building
By the end of this guide you'll have a tested, reusable prompt template for something you actually write or ask for regularly — proven better through a real before-and-after, not a list of tips you're hoping to remember. Anyone can type a question into ChatGPT. This guide teaches something more useful: how to brief AI clearly, judge its first answer honestly, and improve it on purpose.
Get Set Up
Create your ChatGPT account
Go to chatgpt.com and sign up for a free account — using Google, Apple, or your phone number is the quickest way.
The free plan is enough for this entire guide — there's no message limit on regular chat that would get in your way.
You're signed in and see a chat box, ready to type a message.
Stuck?
If you're asked to verify an email or phone number, that's normal account setup — not a sign you need a paid plan.
A quick note on ChatGPT's free plan
As of mid-2026, OpenAI removed the message limit on regular text chat for every plan, including Free — so you won't hit a cap during this guide, even across several rounds of back-and-forth.
The Model: Goal → Context → Format → Iterate
Before you write anything, it helps to see the shape of a prompt that actually works — four pieces, not one clever sentence.
Goal: what you're actually trying to achieve
Not just the surface question — the real outcome you want. "Write an email" is a task. "Reassure a customer without over-apologizing" is a goal.
Context: what AI can't guess
Who it's for, your situation, and any constraints. Skip this, and AI fills the gap with a generic guess.
Format: the shape you want back
Length, structure, tone — say it directly instead of hoping the default matches what you need.
Iterate: refine, don't restart
The first answer is a draft. Improving it is usually faster and better than writing a whole new prompt from scratch.
Try a Weak Prompt on Purpose
This chapter is deliberately about writing a bad prompt first — you'll see exactly why it falls short before fixing it.
Pick a real task
Think of a real message you actually need to write soon — a work update, a reply to a customer, a message to a landlord, anything real.
Using a real task means the before-and-after you're about to see is something you'll actually use, not a throwaway example.
You have one specific real task in mind, and could describe it in a sentence.
Send a deliberately vague prompt
Type just the bare task, with no extra detail — for example, "Write an email about the delay" — and send it.
This is close to what most people's first prompt actually looks like. The point isn't to do it wrong on purpose for fun — it's to see the real cost of skipping Goal, Context, and Format.
ChatGPT returns a generic-sounding response that could apply to almost any situation.
Diagnose why it's weak
Read the response and name at least two specific things wrong with it — wrong tone, missing specifics, wrong length, phrasing that could apply to anyone.
Naming exactly what's wrong is what makes your next prompt a precise fix instead of just trying again and hoping.
You can point to at least two specific problems in the response, not just a general feeling that it's "not quite right."
Rebuild It With the Framework
Now fix it — one piece of the framework at a time, so you can see what each one actually changes.
Add your real Goal
Rewrite your prompt to state what you're actually trying to achieve, not just the task — for example, not "write an email" but "write an email that reassures a customer their order isn't lost, without over-apologizing."
A clear goal changes the substance of the answer, not just the wording.
The new response's content shifts to match the actual goal, not just the surface task.
Add Context AI can't guess
Add who it's for, your situation, and any constraints — for example, the customer's name, what actually happened, and anything you can't promise.
This is the detail that turns a generic answer into one that fits your actual situation.
The response now includes specifics that weren't in your first vague prompt.
Specify the Format
State the format you want directly — for example, "keep it under 100 words, no bullet points, friendly but professional."
Format instructions are the cheapest, most reliable lever you have — AI follows them well when you actually state them.
The response matches the length, structure, and tone you asked for.
Put them side by side
Scroll up to your very first vague response and compare it directly to this final version.
Seeing them side by side — not just remembering the first one felt off — is what makes the improvement concrete instead of a vague impression.
You can point to specific sentences that changed because of Goal, Context, or Format specifically.
Iterate Instead of Restarting
Ask for one refinement
Instead of writing a new prompt, ask for one specific change to the current response — for example, "make this shorter" or "make the tone warmer."
Iterating keeps everything that already worked and changes only what didn't. Restarting throws all of that away and starts the guessing over.
The response updates while keeping the parts that were already good.
Do it again
Ask for one more small refinement, based on what's still not quite right.
Two or three small rounds of iteration almost always beat one attempt at a perfect prompt.
After two rounds of iteration, the response is close to something you'd actually send.
Save It as a Reusable Template
Turn what you just built into something you can reuse — without losing your final result.
Turn your final prompt into a fill-in-the-blank template
Take your final, working prompt and replace the specific details with brackets — for example, "[recipient]", "[what happened]", "[main point]" — keeping the Goal, Context, and Format parts intact.
You're generalizing the specifics, not the structure — the framework is exactly what made it work, so it should stay.
You have a short, reusable template with 3-5 bracketed spots, still recognizably built from Goal, Context, and Format.
Create a Project to keep things organized (optional)
If it's helpful, click "New project" in the ChatGPT sidebar and name it something like "Prompt Templates."
A Project is just a folder for related chats — useful for keeping templates in one place, but it doesn't store or apply anything by itself.
A new project appears in your sidebar, if you chose to create one.
Save your actual template in a chat
Start a new chat — inside your Project, or anywhere — and paste your finished template into it as the first message. ChatGPT automatically titles the chat based on what's inside it, so make sure your template's first line clearly names the task (like "Reusable delay-update email template") — that's what you'll search for later using "Search chats" in the sidebar. No Project handy? A notes app works just as well.
This chat (or note) is your saved template, nothing more — you're not asking anything to apply it automatically.
You can reopen this chat or note later and find your exact template, unchanged.
Why this isn't the same as Project instructions
Project instructions are permanent behavioral rules that apply automatically to every chat inside a Project — things like "keep responses concise," "ask for missing context before answering," or "use a professional tone." A reusable prompt template is different: specific wording for one kind of task, that you paste in on purpose when you need it. Keeping them separate matters — put a single-use template into Project instructions, and it would silently apply to every chat in that Project, whether you wanted it to or not.
What You Learned + Where to Go Next
The pattern behind every good prompt
Goal, Context, Format, Iterate — this works for anything you ask an AI to do, not just this one email. The framework is the reusable part, not any single prompt.
Where to use this next
Apply the same template habit to other recurring tasks — meeting notes, social captions, status updates, anything you write more than once.
The same skill, different domains
This is the same underlying skill the other Level185 guides teach in different settings: directing AI intentionally instead of accepting its first attempt — whether that's a prompt, an automation, or an AI-built website.
Open your saved template, fill it in for a second, different real situation, and send it in a new chat — check whether the very first response already sounds like something you could actually use, with no extra back-and-forth needed.
- You found your saved template without rebuilding it from memory
- The first response needed zero or one small refinement, not several
- You can explain in one sentence why it works, naming Goal, Context, or Format
Recommended next