Skip to main content

36 AI Prompt Ideas You Can Copy and Use at Work

Most prompt lists give you one-line requests that return one-line answers. The prompts below are the ones our team keeps going back to. Each one tells the model who it is helping, what it has to work with and what shape the answer should take. Copy one, swap the parts in brackets, and edit from there.

Key takeaways

  • A good prompt has five parts: role, context, task, format and limits.
  • Pasting your own material in beats describing it. The model can only use what it sees.
  • Ask for a plan or questions first on anything important, then ask for the draft.
  • Treat the first answer as a draft. One follow-up usually doubles its quality.

What makes an AI prompt idea worth using

An AI prompt is the instruction you give a model such as ChatGPT, Claude or Gemini. The model has no idea who you are, who you are writing for, or what you tried last week. Everything it knows about your situation comes from the prompt. That is why "write a blog post about email marketing" produces something you could have found anywhere.

The prompt ideas in this guide share a structure. They give the model a role, hand it real context, ask for one clear task, and say what the output should look like. They also set limits, such as length or things to avoid. You do not need all five parts every time, but when an answer disappoints, one of them is usually missing.

The examples work in any of the major chat assistants. Where a prompt needs your own material, such as a draft, a spreadsheet export or meeting notes, paste it in below the prompt. Do not paste customer data or anything confidential into a tool your company has not approved.

A simple formula for writing your own prompts

Before the list, here is the pattern behind it. If you remember only this table, you can write your own prompt for almost any task.

PartWhat it tells the modelExample
RoleWhose judgement to applyYou are an editor for a B2B software blog.
ContextThe facts it cannot guessOur readers are operations managers at logistics firms.
TaskThe one thing you want doneRewrite the introduction below so the main point comes first.
FormatThe shape of the answerGive two versions, each under 80 words.
LimitsWhat to avoid or respectKeep our product name. Do not add statistics.

AI prompt ideas for marketing and content

These are for planning and editing. We rarely publish anything a model wrote start to finish, but it is very good at finding gaps and giving you options to react to.

  1. Interview me before you write

    Use it when you know the topic well but the page is blank.

    You are a content editor. I want to write about [topic] for [audience]. Before you draft anything, ask me the eight questions you most need answered to make this piece specific and useful. Ask them one at a time and wait for my reply. After the last answer, propose an outline.

  2. Content angles from one idea

    Use it when you have a subject but every headline sounds the same.

    Here is a topic: [topic]. Our reader is [who they are and what they are trying to do]. Give me ten different angles for an article. For each one, write a working title, the single claim the piece would argue, and one reason a reader might disagree with it.

  3. Tighten a draft without changing the voice

    Use it when your draft says the right things but takes too long to say them.

    Edit the draft below. Cut it by about [30]% without removing any facts or examples. Keep my wording where it is already clear. Do not add new sentences. After the edited version, list the five largest cuts you made and why. [paste draft]

  4. Social posts from a long piece

    Use it when an article is published and you need to promote it.

    Read the article below. Write [five] LinkedIn posts that each make one point from it. Each post should open with a concrete observation, not a question, stay under [120] words, and end by pointing to the article. Vary the structure so they do not read like a series. [paste article]

  5. Reader objections check

    Use it when you are about to publish a landing page or announcement.

    Act as a sceptical [job title] who is considering [product or offer]. Read the page copy below and list every question, doubt or missing detail that would stop you from taking the next step. Order them from most to least serious. Then suggest one sentence I could add to answer each of the top five. [paste copy]

  6. A content calendar you can actually keep

    Use it when you need a month of topics that support one goal.

    We sell [product or service] to [audience]. Our goal this quarter is [goal]. We can publish [number] pieces a week. Plan four weeks of content in a table with columns for week, topic, format, the search or sales question it answers, and the call to action. Do not repeat a topic in a different format.

AI prompt ideas for SEO

A model cannot see search volumes or your rankings unless you paste them in. Use it for structure and judgement, and bring the data from your own tools.

  1. Search intent for a keyword

    Use it when you are deciding what kind of page a keyword needs.

    The keyword is "[keyword]". Describe the most likely reasons someone searches for this, split into informational, commercial and transactional intent. For each, say what type of page would satisfy it and what the searcher wants to see in the first screen.

  2. Outline from competing pages

    Use it when you have the top results open and want to cover the topic properly.

    Below are the headings from the top [five] pages ranking for "[keyword]". Identify the subtopics all of them cover, the ones only some cover, and anything none of them address that a reader would still want. Then write an outline for a page that covers the common ground and adds the missing parts. [paste headings]

  3. Title and meta description options

    Use it when a page is written and needs a better search snippet.

    Write eight title tags under 60 characters and eight meta descriptions under 155 characters for the page below. The primary keyword is "[keyword]" and should appear near the start of each title. Make the descriptions say what the reader gets, in plain words. No exclamation marks. [paste page summary]

  4. Questions people ask

    Use it when you want an FAQ section that answers real questions.

    List fifteen questions a [type of buyer] would ask about [topic] before purchasing, in the words they would type into a search box. Group them by stage: just learning, comparing options, ready to buy. Mark the five you think are asked most often.

  5. Internal linking suggestions

    Use it when you have a new page and a list of existing URLs.

    Here is a new article and a list of our existing pages with their titles. Suggest which existing pages should link to the new article and which the new article should link to. For each link, give the anchor text and the sentence where it fits most naturally. [paste article] [paste list of pages]

  6. Refresh an old article

    Use it when a post has lost traffic and needs updating.

    This article was published in [year]. Read it and list what is probably out of date, what is thin compared with what a reader would expect today, and what should be removed. Do not rewrite it. Give me a prioritised update plan with the reason for each change. [paste article]

AI prompt ideas for email and sales

Short messages are harder than long ones. These prompts work because they force the model to be specific about the person on the other end.

  1. Cold email that respects the reader

    Use it when you are writing to someone who has never heard of you.

    Write a first-touch email to [name], [role] at [company]. What I know about them: [one or two facts]. What we do: [one sentence]. The email should be under 90 words, mention one specific reason I chose them, make one clear ask, and sound like a person wrote it. No flattery and no "I hope this finds you well".

  2. Follow-up after no reply

    Use it when your first message went unanswered.

    Below is an email I sent [number] days ago with no reply. Write a follow-up under 60 words that adds one new piece of value, such as a relevant example or a shorter ask, instead of repeating the first email. Give me two versions with different angles. [paste email]

  3. Subject lines to test

    Use it when a campaign is ready but the open rate worries you.

    The email below goes to [audience]. Write twelve subject lines under 45 characters. Make four that state the benefit, four that are specific and factual, and four that raise curiosity without being vague. Tell me which two you would test first and why. [paste email]

  4. Discovery call preparation

    Use it when you have a first meeting with a prospect tomorrow.

    I have a 30-minute call with [role] at a [size and type of company]. They reached out about [problem]. Write ten questions that will help me understand their situation, in an order that moves from easy to more probing. Add the three things I should listen for that would mean we are not a good fit.

  5. Turn call notes into a follow-up

    Use it when the meeting is over and you need to send a summary.

    Here are my rough notes from a sales call. Write a follow-up email that thanks them briefly, summarises what they told me in their own terms, lists the next steps with owners and dates, and flags anything I still need from them. Keep it under 180 words. [paste notes]

  6. Handle a common objection

    Use it when the same pushback keeps coming up.

    Prospects often tell us: "[objection]". We sell [product] to [audience]. Give me three honest ways to respond. Each should acknowledge the concern first, then offer a fact, an example or a question. Do not dismiss the objection or use pressure.

AI prompt ideas for data and reporting

This is the work we do every day. A model will not replace checking the numbers, but it is a fast second pair of eyes on logic, queries and wording.

  1. Explain what a table is telling you

    Use it when you have an export and need the story in it.

    Below is a table of [what it contains] for [period]. Describe the three most important things it shows, with the numbers that support each. Then list anything that looks unusual and might be a data problem instead of a real change. Do not speculate about causes you cannot see in the data. [paste table]

  2. Write a SQL query from a plain request

    Use it when you know what you want but not the exact syntax.

    I use [database, for example BigQuery]. Here are the tables and columns: [paste schema]. Write a query that returns [what you need], grouped by [dimension], for [date range]. Add a comment above each step. Then tell me what assumptions you made about the data that I should verify.

  3. Review a query for mistakes

    Use it when a query runs but the numbers feel wrong.

    Review the SQL below. Look for joins that could duplicate rows, filters that might drop records unintentionally, date ranges that are off by one, and anything that would make totals disagree with a simple count. Explain each issue in one or two sentences and show the corrected line. [paste query]

  4. Define a metric properly

    Use it when two teams report different numbers for the same thing.

    We need one written definition for "[metric]". Ask me the questions you need to pin it down: what counts, what is excluded, which date applies, which system is the source, and how edge cases are treated. After my answers, write the definition in a form I can put at the top of a dashboard.

  5. Summary for leadership

    Use it when you need to send the weekly numbers to people who will read two lines.

    Here are this week's figures and last week's. Write a summary for executives in five bullet points. Lead with what changed most, give the number and the comparison, and say plainly whether it needs a decision. No adjectives such as "strong" or "significant" unless a number follows them. [paste figures]

  6. Plan an analysis before running it

    Use it when someone asks a vague question such as why sales dropped.

    A stakeholder asked: "[question]". We have data on [what you have]. Before I pull anything, list the possible explanations, the specific cut of data that would confirm or rule out each one, and the order to check them in so the cheapest tests come first.

AI prompt ideas for everyday work

Meetings, documents and decisions. Small tasks that add up to hours each week.

  1. Meeting notes into actions

    Use it when you have a transcript or messy notes.

    Turn the notes below into three sections: decisions made, action items with an owner and a date for each, and open questions. If an owner or date was not stated, write "not assigned" instead of guessing. Keep each item to one line. [paste notes]

  2. Summarise a long document

    Use it when you have twenty pages and ten minutes.

    Summarise the document below for someone who has to decide [what decision]. Give me the main argument in two sentences, the five facts that matter most for that decision, and anything the document leaves unclear. Quote the exact wording for any figure or commitment. [paste document]

  3. Push back on my plan

    Use it when you want a critic before your manager becomes one.

    Here is my plan for [project]. Act as an experienced [role] who has seen similar projects fail. Tell me the weakest three parts, what would have to be true for each to work, and what you would ask me in a review meeting. Be direct. I do not need encouragement. [paste plan]

  4. Rewrite for a different reader

    Use it when a technical message has to go to a non-technical person.

    Rewrite the message below for [who they are]. They do not know the terms [list jargon]. Keep every fact and every request. Use short sentences and put what they need to do in the first line. [paste message]

  5. Draft a process document

    Use it when a task lives in one person's head.

    I am going to describe how we do [task]. Ask me questions until you understand each step, who does it, what tool they use and what can go wrong. Then write it as a numbered procedure that a new team member could follow without help.

  6. Compare options fairly

    Use it when you are choosing between tools, vendors or approaches.

    I am deciding between [option A], [option B] and [option C] for [purpose]. What matters most to us: [criteria in order]. Build a comparison table using only what I tell you and what is generally known, and mark anything you are unsure about. Finish with the question I should answer to break a tie.

AI prompt ideas for learning and thinking

Use these when you want to understand something, not only get an answer.

  1. Teach me in steps

    Use it when you are new to a subject and textbooks start too far ahead.

    Teach me [topic]. I already know [what you know] and nothing more. Explain one idea at a time, give an example from [a field you know well], then ask me a question to check I understood before moving on.

  2. Explain it three ways

    Use it when one explanation did not land.

    Explain [concept] three times: once in two plain sentences, once with an everyday comparison, and once the way a specialist would say it to a colleague. Then tell me the most common misunderstanding people have about it.

  3. Quiz me

    Use it when you have read the material and want to know if it stuck.

    I have been studying the notes below. Ask me ten questions, one at a time, starting easy and getting harder. After each answer, tell me whether I was right and correct me briefly if not. At the end, list the areas I should review. [paste notes]

  4. Find the gaps in my thinking

    Use it when you have formed an opinion and want it tested.

    Here is what I believe about [subject] and why. Give me the strongest case against my view, as someone who holds the opposite position would argue it. Then tell me which of my reasons is weakest and what evidence would change my mind. [paste your view]

  5. A study plan that fits your week

    Use it when you want to learn a skill alongside a full-time job.

    I want to learn [skill] to the point where I can [specific outcome]. I have [hours] a week for [number] weeks. Build a weekly plan with what to study, one small project per week to practise it, and how I will know I am ready to move on.

  6. Read a paper or report with me

    Use it when a dense document is beyond your background.

    I am going to paste a section of a [research paper or report]. Explain what it says in plain language, define each term the first time it appears, and tell me what a careful reader would question. Keep your explanation shorter than the original. [paste section]

How to adapt a prompt so it works for you

A copied prompt is a starting point. These habits are what turn an average reply into one you can use.

  1. Fill every bracket with something real

    "Marketing managers" is weaker than "marketing managers at 50-person software companies who report to a founder". The more exact the reader and the situation, the less generic the answer.

  2. Paste the source, do not describe it

    If you want a draft edited, a table read or notes organised, put the actual text under the prompt. Summarising it yourself throws away the detail the model needs.

  3. Ask for the format you will use

    A table, five bullets, a 90-word email. If you leave the shape open you will spend time reformatting a long essay.

  4. Follow up at least once

    Say what was wrong: too long, too formal, missed the main point. The second reply is nearly always better, because now the model has your reaction as context.

  5. Check anything that looks like a fact

    Models can state wrong numbers, dates and quotes with complete confidence. Verify figures, names and citations against a source before they leave your hands.

Mistakes that make AI prompts fail

The most common one is asking for too much at once. A prompt that says research, outline, write, and optimise in a single request gets a shallow version of each. Split the work. Ask for the outline, react to it, then ask for a section.

The second is vague quality words. "Make it engaging" or "more professional" means something different to everyone. Say what you want changed: shorter sentences, the conclusion first, fewer adjectives, an example in each paragraph.

The third is treating the answer as finished. A model writes what is plausible, which is not always what is true or what fits your business. The people who get the most from these tools use them to produce a first draft or a second opinion, and keep the final judgement for themselves.

AI prompt ideas: frequently asked questions

What is an AI prompt?

An AI prompt is the text you give an AI model to tell it what you want. It can be a question, an instruction, or an instruction plus material such as a document or data for the model to work with.

What are good AI prompt ideas for beginners?

Start with tasks where you can judge the result easily: summarising a document you have read, turning meeting notes into action items, or rewriting an email for a different reader. You will see quickly what the model does well and where it needs more context.

How long should a prompt be?

As long as it needs to be to give the model the role, context, task, format and limits. That is often three to six sentences plus any pasted material. Length is not the goal, but one-line prompts usually leave out something the model needed.

Do these prompts work in ChatGPT, Claude and Gemini?

Yes. They are written in plain language and do not depend on features of one product. Answers will differ between models, so try the same prompt in more than one if the first result is not what you wanted.

Is it safe to paste company data into an AI tool?

Only if your company has approved that tool for that kind of data. Check your employer's policy and the provider's data terms first, and remove personal or confidential details when you are unsure.

How do I get less generic answers?

Give more specific context, paste real examples of the style you want, state what to avoid, and ask the model to ask you questions before it writes. Generic input is the main cause of generic output.

Nous Contacter

Êtes-vous prêt à développer votre entreprise ?

Passez votre marketing au niveau supérieur avec les outils alimentés par l'IA de Multivariate.

  • Fast ResponseWithin 24 hours
  • Expert TeamReady to help
  • No ObligationJust a friendly chat
Trusted by 500+ businesses

Nous Contacter

Tell us about your project

We’ll get back to you within 24 hours.

We respect your privacy. No spam, ever.