---
title: "Good Prompts for ChatGPT: 8 Templates You Can Paste Now"
description: "Eight copy-ready prompts for ChatGPT, from meeting notes to resumes and funny games, plus the three things that make any prompt work."
url: https://www.promptpal.app/blog/good-prompts-for-chatgpt-templates
author: "PromptPal Team"
published: 2026-11-15T08:04:00.017Z
updated: 2026-11-21T22:03:05.446Z
category: "General"
tags: ["chatgpt prompts", "prompt templates", "prompt writing", "resume prompts", "prompt library"]
language: en
reading_time_minutes: 9
image: https://images.ctfassets.net/xzkuxcljsk12/6MyaxZ5mlvZC1Bs2ZZeh1e/4d229df2d2ffc075b5cd21d7580c5ffc/blog-1.png
publisher: PromptPal
---

# Good Prompts for ChatGPT: 8 Templates You Can Paste Now

Eight copy-ready prompts for ChatGPT, from meeting notes to resumes and funny games, plus the three things that make any prompt work.

A good prompt for ChatGPT states the task, gives the context the model can't guess, and says what the answer should look like. Everything below follows that pattern. Copy the templates, fill in the `[placeholders]`, and keep the ones that earn a permanent spot.

The list is short on purpose. Eight prompts that cover most of a working week beat a hundred you will never open.

## What makes a prompt good (the short version)

Three things separate a prompt that works from one that produces mush: a clear job, real context, and a stated output format. Leave out any one and ChatGPT fills the gap with the most average guess available.

Here is a before and after for a task everyone does.

**Before:**

> `Write an email asking a client to pay an overdue invoice.`

You get a stiff, generic email that apologises three times and could have come from any company. It has no invoice number, no tone, no deadline.

**After:**

> `Write a short email to a client whose invoice is overdue.`
>
>
>
> `Context:`
>
> `- Invoice [number] for [amount] was due on [date]`
>
> `- We've worked with them for [length of time] and the relationship is good`
>
> `- This is the first reminder`
>
>
>
> `Tone: friendly and direct, not apologetic and not threatening.`
>
> `Length: under 120 words.`
>
> `End with one clear ask: payment by [new date], or a reply if there's a problem.`
>
> `Give me a subject line too.`

The second version returns something you can send after a quick read. The answer changed because the model no longer had to invent the relationship, the tone or the ask. That is the whole trick, and the templates below are that trick applied to common jobs.

Notice what the better prompt leaves out. There is no "you are a world-class accounts manager" and no "be professional." Those lines feel like effort but add almost nothing. Facts, limits and a defined ending do the work.

If you want the longer explanation of why this works, our [guide to writing better ChatGPT prompts](https://promptpal.app/blog/how-to-write-better-chatgpt-prompts-a-complete-guide-2026) covers it.

## Eight good prompts for ChatGPT

Each one is complete as written. Paste it, replace the brackets, send.

### 1. Summarise a meeting into actions

This is the prompt most people end up using weekly. The key is forcing it to separate decisions from tasks, because a plain "summarise" blends them into one paragraph nobody acts on.

> `You're helping me write up a meeting. Below are the raw notes or transcript.`
>
>
>
> `Return:`
>
> `1. Decisions made (one line each)`
>
> `2. Action items as a list: owner, task, due date. If the owner or date wasn't stated, write "unassigned" or "no date" instead of guessing.`
>
> `3. Open questions nobody answered`
>
>
>
> `Keep it under 200 words. Don't add anything that isn't in the notes.`
>
>
>
> `Notes:`
>
> `[paste notes or transcript]`

The "don't guess" line matters. Without it, ChatGPT happily assigns tasks to people who never agreed to them.

Here is what that looks like on messy input. Say your notes include: "Priya thinks we should move the launch, Sam disagrees, ended up keeping the 14th. Someone needs to tell legal. Pricing page copy still not done, who's on that?" A good run returns one decision (launch stays on the 14th), one action (tell legal, owner unassigned, no date), one action (pricing page copy, unassigned) and one open question. It does not quietly hand the legal task to Priya. That small refusal to invent is the difference between a summary you trust and one you have to re-check line by line.

### 2. Rewrite your own draft without losing your voice

> `Edit the text below. Fix clarity and cut filler, but keep my voice: [describe your voice, e.g. blunt, dry, first-person].`
>
>
>
> `Rules:`
>
> `- Don't add new claims or examples`
>
> `- Keep paragraphs short`
>
> `- Tell me what you cut and why, in three bullets at the end`
>
>
>
> `Audience: [who will read this]`
>
> `Text:`
>
> `[paste draft]`

Asking for the three-bullet explanation catches the moments when the edit drifts away from what you meant. If the bullets say "removed your joke in paragraph two," you know to put it back.

### 3. Explain something at the right level

> `Explain [topic] to someone who [describes the reader, e.g. is a marketing manager with no technical background].`
>
>
>
> `Start with the one-sentence version. Then give a concrete example from [their field]. Then list two things people commonly get wrong.`
>
> `Avoid jargon; if you must use a term, define it in the same sentence.`

Naming the reader is what changes the answer. "Explain it simply" gives you a different, usually more patronising, result. The example from the reader's own field is the part that makes it stick.

### 4. Compare options and make a call

> `I'm choosing between [option A] and [option B] for [purpose].`
>
>
>
> `My priorities, in order: [priority 1], [priority 2], [priority 3].`
>
> `Constraints: [budget, deadline, team size, anything fixed].`
>
>
>
> `Compare them against my priorities only, not generic pros and cons. Then recommend one and say what would change your mind. If you need information from me before you can recommend, ask up to three questions first.`

Without stated priorities you get a balanced list that helps nobody decide. With them, you get an actual recommendation you can argue with, which is the point. The "what would change your mind" line also exposes how shaky the recommendation is.

### 5. Resume bullet rewrite

Resume prompts for ChatGPT go wrong in one predictable way: the model invents numbers and achievements. It does this because strong resume bullets in its training data nearly always contain a metric, so it supplies one. This version blocks that.

> `Rewrite the resume bullets below for a [job title] application.`
>
>
>
> `Rules:`
>
> `- Use only facts I've given you. Do not invent metrics, tools, team sizes or outcomes.`
>
> `- If a bullet would be stronger with a number I haven't provided, leave a [placeholder] and tell me what to fill in.`
>
> `- Start each bullet with a strong verb. One line each.`
>
>
>
> `Job description (key requirements only):`
>
> `[paste]`
>
>
>
> `My current bullets:`
>
> `[paste]`

Paste only the key requirements from the job ad, not the whole thing. Then read each rewritten bullet against your original and delete anything you can't defend in an interview. We wrote a longer set for Claude, and the templates carry over to ChatGPT: [resume prompts that don't invent](https://promptpal.app/blog/resume-prompts-for-claude).

### 6. Critique before you ship

> `Act as a sceptical reviewer of the [document type, e.g. proposal, landing page, cover letter] below. The goal is [what it should achieve] and the reader is [who].`
>
>
>
> `Give me:`
>
> `- The three weakest points, ranked, with a quote from the text for each`
>
> `- One thing that's working and should stay`
>
> `- A rewrite of the single weakest paragraph`
>
>
>
> `Don't soften the feedback.`
>
>
>
> `Text:`
>
> `[paste]`

By asking for quotes, you force it to point at real lines instead of offering vague praise. Run it once before you send anything that matters, such as a proposal or a cover letter.

### 7. Fun prompts for ChatGPT: a game you can actually play

Fun and funny prompts for ChatGPT work best when they have rules. "Be funny" gets you puns. A constraint gets you something worth reading aloud.

> `Let's play a game. You're a [character, e.g. a weary medieval innkeeper] reviewing [everyday thing, e.g. my team's weekly stand-up].`
>
>
>
> `Rules:`
>
> `- Stay in character the whole time`
>
> `- Write exactly five lines, each one a complaint more dramatic than the last`
>
> `- End with a one-star rating and a single sentence of reluctant praise`

Swap the character and the subject and you have a hundred variations: a nature documentary narrator describing your commute, a noir detective investigating a missing stapler. The structure (character, subject, line count, ending) is what makes it land. Humour is also the one area where you should expect to regenerate a few times and pick the best.

### 8. Turn a messy idea into a plan

> `Here's a rough idea: [describe it in a few sentences, mess included].`
>
>
>
> `Turn it into a plan:`
>
> `1. Restate the goal in one sentence so I can confirm you understood`
>
> `2. List the steps in order, each with a rough effort level (small, medium, large)`
>
> `3. Flag the two assumptions most likely to be wrong`
>
> `4. Suggest the smallest first step I could finish today`
>
>
>
> `If the idea is too vague to plan, ask me questions instead.`

The restatement step catches misunderstandings before you read a whole plan built on the wrong goal.

## Five habits that quietly make prompts worse

Most bad ChatGPT answers trace back to the prompt, and the same few habits show up again and again.

- **Stacking tasks.** "Summarise this, translate it, and suggest a headline" gets three mediocre results. Run them as separate steps, or number them and ask for each in order.
- **Vague adjectives.** "Professional" and "engaging" mean nothing specific. "Plain, no exclamation marks, no sentences over 20 words" is something it can follow.
- **Pasting without labelling.** If you drop in a block of text with no note on what it is, the model has to guess whether it's a draft, a source or an example. Say which.
- **No way out.** When information is missing, the model fills it in. Tell it to ask or to leave a placeholder.
- **Accepting the first answer.** The first reply is a draft. The second message, where you say what's off, is usually where the good version appears.

## How to adapt any template

You rarely need to rewrite a prompt from scratch. Change these four things and it fits a new job:

- **The role or reader.** Who is it for, or who is the model standing in as?
- **The facts.** Replace every bracket with real detail. Vague inputs give vague outputs.
- **The limits.** Word count, tone, things to avoid. One or two are enough.
- **The escape hatch.** Tell it what to do when information is missing, such as ask a question or leave a placeholder. This is what stops invention.

If an answer is close but not right, don't start over. Say what's wrong in one line ("too formal, and the second point repeats the first") and ask for a revision. That's faster than rewriting the prompt, and it shows you which part of the original was underspecified. Once you know, fix that line in the template so the next run starts better.

For tasks where you want a certain style, pasting two examples of what you like often does more than a paragraph of description. This is called few-shot prompting, and it works especially well for tone, formatting and anything you struggle to put into words.

Keep prompts shorter than you think they need to be. A template with eight clear lines gets used. A template with forty lines of rules gets skipped, and models often follow the later rules less reliably anyway, though that varies by model and version.

## Which prompts are worth keeping

Most prompts you write are one-offs. A few become part of how you work, and those deserve a home better than a notes file or a chat you'll never scroll back to.

A simple rule: if you've pasted the same prompt twice, save it. Name it by the job ("Meeting notes to actions"), not by the date. Keep the brackets in so it's ready to reuse. When you improve it, keep the old version until the new one has proven itself on a real task.

This is where PromptPal helps. You can save prompts to your own [prompt library](https://promptpal.app/features/prompt-library), group related ones into collections, and refine a draft with the Chrome extension while you're inside ChatGPT. If you work with others, shared workspaces keep everyone on the same version of the prompt instead of five slightly different copies.

One caution on "best prompts for ChatGPT" lists in general, including this one. A prompt that's perfect for someone else's job is a starting point for yours. Test it on a real task of your own, adjust it, and only then save it. Results also vary between models and versions, so a prompt tuned for ChatGPT may need tweaks in Claude or Gemini.

## Next step

Pick the one template above that matches something you do every week, run it on real material, and save the version that works. You can browse ready-made templates by role and task in the [PromptPal public prompt library](https://promptpal.app/prompts) and build your own set from there.

---

Originally published at https://www.promptpal.app/blog/good-prompts-for-chatgpt-templates
