AI Prompt Library
Browse a curated library of prompt templates you can copy in one click and adapt to your task, organised by what you are trying to do: write, learn, code, plan, or get better answers. Each one uses the structure that reliably gets more from ChatGPT, Claude and Gemini.
Rewrite for a different tone
WritingTurn a stiff or rambling draft into the exact voice you need.
You are an experienced copy editor. Rewrite the text below so it sounds [warm and friendly / crisp and professional / confident and direct], while keeping the meaning and all facts unchanged. Match the reading level of a general audience, prefer short sentences, and cut filler. Return only the rewritten version. Text: """ [paste text] """
Shorten without losing meaning
WritingTighten a bloated email, bio or paragraph.
Condense the text below to roughly [50%] of its length. Keep every key point, name and number; remove repetition, hedging and filler. Preserve the original tone. Return only the shortened text, then a one-line note of anything important you had to cut. Text: """ [paste text] """
Draft an email from bullet points
WritingGo from scattered thoughts to a send-ready email.
You are writing on my behalf. Turn the notes below into a [short, polite] email to [recipient and their role]. The goal is to [what you want to happen]. Keep it under [120] words, use a clear subject line, and end with a specific call to action. Notes: - [point 1] - [point 2] - [point 3]
Fix grammar and flow only
WritingClean up writing while keeping it unmistakably yours.
Proofread the text below. Correct grammar, spelling, punctuation and awkward phrasing, but do not change my voice, add new ideas or restructure the argument. Preserve any technical terms exactly. Return the corrected text, then a short bulleted list of the changes you made and why. Text: """ [paste text] """
Brainstorm titles and hooks
WritingEscape a blank page with ten angles to choose from.
Act as a headline writer. Based on the summary below, give me [10] title options for a [blog post / video / newsletter] aimed at [audience]. Vary the angle: some curiosity-driven, some benefit-driven, some contrarian. Keep each under [60] characters and avoid clickbait that overpromises. Summary: [what the piece is about]
Outline before I write
WritingGet a solid skeleton so the drafting goes fast.
You are a structural editor. Create a detailed outline for a [800]-word [article] on [topic] for [audience]. Give me an H1, 4–6 section headings, and 2–3 bullet points under each covering what to say. Note where an example, statistic or story would strengthen the piece. Do not write the full draft yet.
Turn meeting notes into action items
Work & productivityNever let a to-do slip out of a messy meeting again.
You are a diligent chief of staff. From the raw meeting notes below, produce: (1) a 3-sentence summary, (2) a table of action items with columns Owner, Task and Due date, and (3) any open questions or decisions still pending. If an owner or date is missing, write "unassigned" rather than guessing. Notes: """ [paste notes] """
Summarise a long document
Work & productivityGet the gist of a report, contract or thread in seconds.
Summarise the document below for [a busy executive / a new team member]. Give me: a one-sentence TL;DR, the 5 most important points as bullets, and any risks, deadlines or numbers I should not miss. Stay strictly faithful to the source — do not add outside information. Document: """ [paste text] """
Prep for a difficult conversation
Work & productivityWalk into a tense chat with a plan, not just nerves.
You are a communication coach. I need to [raise a concern / give feedback / say no] to [person and relationship]. The situation is: [context]. My goal is [outcome], and I want to stay [calm and respectful]. Give me: an opening line, 3 key points to make, likely objections with how to respond, and one thing to avoid saying.
Plan a project into steps
Work & productivityConvert a vague goal into a sequenced, doable plan.
Act as a project planner. My goal is [describe the goal] by [deadline]. I have [time / resources available]. Break this into a phased plan with milestones, list the tasks under each phase in order, flag dependencies (what must finish before what), and highlight the two biggest risks with a mitigation for each.
Reply to a tricky message
Work & productivityHandle awkward emails without overthinking every word.
Help me reply to the message below. My relationship to the sender is [role], and I want to come across as [professional and helpful] while [holding my position / declining / buying time]. Draft a reply under [100] words, then give me one slightly firmer alternative. Message: """ [paste message] """
Compare options in a table
Work & productivitySee a decision laid out clearly before you commit.
I'm deciding between these options: [option A], [option B], [option C], for the purpose of [goal]. Build a comparison table with the criteria that matter most for this decision (you choose sensible ones and explain them). Then give a one-paragraph recommendation for someone who prioritises [your top priority], noting what would change your answer.
Explain it like I'm five
LearningBreak through a concept that keeps going over your head.
Explain [concept] to me as if I'm five years old. Use a simple everyday analogy, avoid all jargon, and keep it to about [4] sentences. Then add one line that says "In grown-up terms:" and give the precise definition so I can level up when I'm ready.
Teach me step by step
LearningLearn something new at a pace you can actually follow.
You are a patient tutor. Teach me [skill or topic] from scratch. Start by asking what I already know, then explain in small steps, checking my understanding with a quick question after each step before moving on. Use concrete examples. Don't dump everything at once — we'll go one step at a time.
Quiz me on what I'm studying
LearningTurn passive notes into active recall practice.
Act as a study coach. Based on the material below, ask me [8] questions one at a time, mixing recall and application. After each answer, tell me if I'm right, explain briefly, and adjust the difficulty based on how I did. At the end, summarise my weak spots and what to review. Material: """ [paste notes] """
Summarise and find the key idea
LearningGet to the heart of a dense reading fast.
Read the passage below and give me: (1) the single core idea in one sentence, (2) three supporting points, and (3) one thing the author assumes but doesn't prove. Then suggest one question I should ask to understand it more deeply. Passage: """ [paste text] """
Give me analogies for a hard idea
LearningSee the same idea from three angles until it clicks.
I'm struggling to grasp [concept]. Give me three different analogies for it — one from everyday life, one from nature, and one from sport or games. For each, explain exactly where the analogy holds and, importantly, where it breaks down so I don't over-apply it.
Build me a learning plan
LearningReplace random tutorials with a real curriculum.
You are a curriculum designer. I want to learn [topic] and I can spend [hours per week] for [number of weeks]. My current level is [beginner]. Create a week-by-week plan with clear goals, free resources to look for, and a small project each week to practise. End with how I'll know I've actually learned it.
Debug this code
CodingFind the bug faster than staring at it will.
You are a senior [language] developer. The code below is [not working as expected / throwing an error]. Expected behaviour: [what should happen]. Actual behaviour: [what happens, including any error message]. Explain the likely cause, show the corrected code, and point out anything else risky you notice. Code: ``` [paste code] ```
Explain this code line by line
CodingUnderstand unfamiliar code before you touch it.
Explain the code below to someone who knows [language] basics but not this codebase. Walk through what it does at a high level first, then annotate the tricky parts. Flag anything that looks like a bug, a performance issue or a bad practice. Don't rewrite it — just explain. Code: ``` [paste code] ```
Write a function to a spec
CodingGet a clean, tested function instead of a rough sketch.
Write a [language] function named [name] that [does what]. Inputs: [describe]. Output: [describe]. Constraints: [edge cases, performance, no external libraries, etc.]. Include brief comments, handle invalid input gracefully, and add 3 example test cases showing it works. Explain any assumptions you made.
Review my code
CodingCatch problems before they reach production.
Act as a code reviewer. Review the code below for correctness, readability, security and performance. List issues grouped by severity (blocker / should-fix / nice-to-have), each with a short explanation and a suggested fix. Be direct but constructive, and call out anything done well too. Code: ``` [paste code] ```
Turn an error message into a fix
CodingDecode a cryptic stack trace into next steps.
I got this error while [what you were doing]. Environment: [language / framework / version]. Here's the full error and stack trace: ``` [paste error] ``` Explain in plain terms what the error means, the most likely causes ranked by probability, and the exact steps to fix it. If you need more information to be sure, tell me what to check.
Write a regular expression
CodingGet a regex you can actually understand and trust.
Write a regular expression that matches [describe exactly what should match]. It should also NOT match [describe what to exclude]. Target flavour: [JavaScript / Python / PCRE]. Show the pattern, explain each part, and give 3 examples that match and 3 that don't so I can verify it.
Plan meals for the week
Everyday lifeSolve 'what's for dinner' for a whole week at once.
You are a practical meal planner. Create a [7]-day dinner plan for [number] people. Dietary needs: [vegetarian / allergies / preferences]. Budget-conscious and each meal should take under [30] minutes. Give me the plan as a day-by-day list, then a consolidated shopping list grouped by supermarket section.
Plan a trip itinerary
Everyday lifeGet a realistic, well-paced plan instead of a wish list.
Act as a local travel guide. Plan a [3]-day trip to [destination] for [who's going] with interests in [food / history / nature / nightlife]. Budget level: [mid-range]. Give a day-by-day itinerary that groups things by area to save travel time, mixes must-sees with a couple of local secrets, and leaves some breathing room. Note the best time of day for each.
Draft a tricky personal message
Everyday lifeSay the hard thing well when the words won't come.
Help me write a [text / message] to [person and relationship]. I need to [apologise / decline an invite / set a boundary / check in]. The context is: [situation]. I want to sound [sincere and kind] without being [over-the-top]. Keep it natural and human — the way a real person texts, not a form letter. Give me two versions of different lengths.
Break a big goal into small steps
Everyday lifeTurn an intimidating goal into a doable first move.
I want to [big personal goal] but it feels overwhelming. Break it into a realistic plan of small steps I can actually start this week. Assume I have [time / energy / constraints]. Make the first step tiny enough that I can't say no, and suggest a simple way to track progress so I stay motivated.
Explain a bill, form or contract
Everyday lifeUnderstand the fine print before you sign it.
Explain the document below to me in plain language. Tell me what it's actually saying, what I'm agreeing to or being charged, anything unusual or that favours the other party, and any questions I should ask before I sign or pay. You are not a lawyer, so flag where I should get professional advice. Document: """ [paste text] """
Get unstuck on a decision
Everyday lifeThink a personal decision through with a patient sounding board.
I'm trying to decide whether to [decision]. Here's my situation: [context], and what matters most to me is [values / priorities]. Don't just list pros and cons — ask me [3] clarifying questions first, then help me think it through and reflect back what your read on my leaning is. I'll make the final call.
Think step by step (chain of thought)
Better answersGet more reliable answers on anything with reasoning.
Solve the problem below. Think step by step, showing your reasoning for each step before you give the final answer. If there's more than one way to interpret the question, state your assumption first. End with a clearly labelled "Answer:" line. Problem: """ [paste problem] """
Give examples first (few-shot)
Better answersShow, don't tell — the fastest way to nail a format.
I want you to [task] in a specific style. Here are examples of the input and the output I want: Input: [example input 1] Output: [example output 1] Input: [example input 2] Output: [example output 2] Now do the same for this input, matching the style and format of the examples exactly: Input: [your real input]
Assign a role and audience
Better answersPrime the model to answer like the right kind of expert.
You are [a specific expert role, e.g. a tax accountant / a UX researcher / a fitness coach] with [years] of experience. You're explaining to [audience and their level]. Your task is to [what you want]. Prioritise [accuracy / practicality / simplicity], and if something is uncertain or depends on context, say so rather than guessing.
Critique my reasoning
Better answersPressure-test an idea before reality does.
Play devil's advocate. Below is my argument / plan. Find the weakest points, the assumptions I haven't examined, and the strongest counterargument someone could make. Be rigorous and honest, not agreeable — I want to find the holes now. Then suggest how I could strengthen it. Argument: """ [paste your reasoning] """
Ask me questions before answering
Better answersStop the model guessing and get a tailored answer.
Before you answer, ask me any clarifying questions you need to give me a genuinely useful response to: [your request]. Don't assume — if key details are missing (audience, goal, constraints, format), ask for them first. Once I answer, then give me your best response.
Specify the exact output format
Better answersGet output you can paste straight into a doc or app.
[Your task]. Return the result strictly as [a markdown table with columns X, Y, Z / a JSON object with these keys / a numbered list of no more than 5 items]. Do not include any preamble, explanation or closing remarks — output only the requested format. If you can't fill a field, use "n/a".
Rate your own confidence
Better answersKnow when to trust the answer and when to double-check.
Answer the question below. Then, on a separate line, rate your confidence in the answer as High, Medium or Low and explain why in one sentence. If your confidence is Low or the answer could be out of date, say what I should verify independently. Question: [your question]
Iterate: make it better
Better answersRefine a good-but-not-great answer instead of starting over.
Here is your previous response: """ [paste the answer] """ Now improve it. Specifically: [make it shorter / more concrete / less formal / add examples / fix X]. Keep what's working, change only what I asked, and briefly note what you changed at the end.
The anatomy of a prompt that works
Most disappointing AI answers trace back to the same cause: the prompt didn’t give the model enough to work with. A strong prompt isn’t a magic incantation — it’s clear communication with five moving parts. Get these right and the quality of your answers jumps, no matter which assistant you use.
- Role. Tell the model who to be. “You are an experienced copy editor” or “Act as a patient tutor” primes it to draw on the right knowledge and adopt the right voice.
- Context.Give the background it can’t guess: who the audience is, what you’ve already tried, the situation, the source text. This is where most weak prompts fall short.
- Task. State plainly what you want done. One clear task beats three vague ones bundled together.
- Format. Specify the shape of the answer: a table, a numbered list of five, a JSON object, an email under 120 words. Models will happily oblige when you ask.
- Constraints.Set the guardrails: tone, length, reading level, what to avoid, and what to do when it’s unsure.
You don’t need every element every time — a quick question needs none of this. But whenever an answer matters, walking through these five parts is the fastest way to a better result. The templates above are built on exactly this skeleton, which is why they hold up.
Techniques worth knowing
A handful of techniques do most of the heavy lifting in prompting. None of them are complicated, and they stack — you can combine them in a single prompt.
Give examples (few-shot prompting)
The most reliable way to get a specific style or format is to show it. Paste two or three examples of the input and the output you want, then give the real input. This is called few-shot prompting, and examples communicate far more precisely than any description. If the model keeps missing your intended format, add an example — it almost always fixes it.
Ask it to think step by step (chain of thought)
For anything involving reasoning — maths, logic, multi-step planning, tricky judgement calls — adding “think step by step and show your reasoning” measurably improves accuracy. This is known as chain-of-thought prompting. It works because it gives the model room to work the problem out rather than blurting a first guess.
Assign a role
Naming a persona — a tax accountant, a UX researcher, a fitness coach — nudges the model toward the vocabulary, priorities and standards of that field. Pair the role with the audience (“explaining to a beginner”) and the answer calibrates itself to the right level.
Iterate
The first answer is a draft, not a verdict. The fastest path to a great result is often to accept a decent one and then refine: “shorter,” “more concrete,” “less formal,” “add an example.”Treat the conversation as a collaboration and you’ll get further than trying to write one perfect prompt up front.
Common mistakes
If your answers are underwhelming, the fix is usually in the prompt. These three mistakes account for most of the frustration.
- Too vague. “Write me something about marketing” could mean a thousand things, so you get a bland average of all of them. Add the audience, the goal, the format and the length.
- Too much at once. Cramming a research task, a draft and a rewrite into one prompt makes the model juggle and drop things. Break big jobs into steps and go one at a time.
- No format specified.If you don’t say how you want the answer, you’ll get a wall of prose you then have to reshape by hand. Ask for the table, the bullets or the JSON up front.
A subtler mistake is assuming the model knows things it can’t. It hasn’t read your earlier chats, your files or your mind unless you include them. When in doubt, paste the context in — or use a prompt that asks the model to request whatever it’s missing before it answers.
How to adapt these templates
Every prompt above is a starting point, not a script. The bracketed placeholders — [topic], [audience], [paste text]— are the parts you replace with your own details. But don’t stop there: change the tone, tighten the length, add a constraint, or merge two templates if your task needs it. The best prompt is the one shaped around your actual situation.
Copy a template, fill in the brackets, run it, and read the answer critically. If it missed the mark, you now know exactly what to adjust — add an example, specify the format more tightly, or give it the context it was clearly guessing at. A minute of iteration beats hunting for a “perfect” prompt someone else wrote.
One more thing: the right model matters as much as the right prompt. The same prompt can land very differently in a fast, cheap model versus a top-tier reasoning model. If you’re not sure which assistant fits your task, our Which AI should you use? tool gives a quick recommendation, and our AI model comparison lays out the strengths side by side. New to the vocabulary in these prompts? The AI glossary has plain-English definitions.
Frequently asked questions
What is prompt engineering?
Prompt engineering is the craft of writing instructions that get the best out of an AI model. In practice it means being specific about the role, context, task, format and constraints — so the model has everything it needs to give a useful answer. It is less about secret tricks and more about clear communication, and a few reliable structures do most of the work.
What makes a good prompt?
A good prompt tells the model who it should act as, gives the background it needs, states the task plainly, specifies the exact output format, and sets any constraints (length, tone, what to avoid). Vague prompts get vague answers; specific prompts get specific ones. Adding one or two examples of what you want is one of the most effective upgrades you can make.
Do these prompts work in ChatGPT, Claude and Gemini?
Yes. Every template here is written in plain language and avoids tool-specific syntax, so you can paste it into ChatGPT, Claude, Gemini, Copilot or most other assistants and it will work. Results vary a little between models because each has different strengths, but the structure of a good prompt is universal.
What is few-shot prompting?
Few-shot prompting means giving the model a few worked examples of the input and the output you want before asking it to do the real task. It is one of the most powerful techniques because examples communicate a style or format far more precisely than a description can. Giving zero examples is called zero-shot; one is one-shot; a handful is few-shot.
Are the prompts free to copy?
Yes — every prompt in this library is free to copy, paste, edit and use, at work or at home, no attribution required. They are templates meant to be adapted: swap the bracketed placeholders for your own details and reshape them however you like.
Does anything I type here get sent anywhere?
No. This library runs entirely in your browser. Searching and copying happen on your device, nothing you type is uploaded, logged or stored, and the prompts themselves are just text. What you do with them inside ChatGPT, Claude or Gemini is governed by that provider's own privacy policy.
Actually understand AI, one minute a day.
Scroll: Learn AI turns everything behind these tools into bite-sized lessons and quizzes. Free on iOS, Android coming soon.
