200+ ChatGPT Prompts: The Copy-Paste Library for 2026

A prompt library is only worth having if the prompts are specific. Most lists that circulate under this title are a thousand variations on write me a blog post about [topic], which is the prompt equivalent of asking a contractor to do some building. The 216 prompts below are grouped by the job they do, and each one carries the context, constraint or output format that separates a usable first draft from a paragraph you throw away.

Everything here is copy-paste ready. Square brackets mark the parts you replace. Read the first two sections before you scroll to the library itself, because they cover the two things that actually determine what you get back: the shape of the prompt, and the fact that the same prompt run twice will not give you the same answer.

What separates a prompt that works from one that does not

Four elements do most of the work: a role, the context, the task, and the output format. Role sets the vocabulary and the assumed standards. Context is the material the model would otherwise invent. Task is the single thing you want done. Format is how you want it handed back, and it is the element people skip most often, which is why so many answers arrive as an essay when a table would have been useful.

We took that skeleton apart in more detail, with worked examples, in how to write ChatGPT prompts. The short version is that structure reliably controls the shape of the answer. It does not control the substance, which brings us to the caveat that matters most.

The same prompt will not give you the same answer twice

This is the part prompt libraries tend to leave out. Answer engines are not deterministic, and the gap between runs is wide enough to change a decision. SparkToro found that the same query changes its answer roughly 70% of the time, and that two identical queries return the same list of recommended brands less than once in a hundred attempts.

Our own measurement puts a number on the sharp end of that. Across 20 buyer-intent prompts run three times each through the engines' APIs on 13 July 2026, the number-one recommended brand changed between consecutive identical runs 44% of the time on Gemini and 28% of the time on Claude.

Two practical consequences. First, if a prompt produces something good, save the prompt and the output, because you may not reproduce it. Second, never judge anything important on a single run, whether that is a research answer or a check on how an engine describes your company. We covered the reproducibility problem for consumers in whether ChatGPT gives everyone the same answer.

Writing and editing

• You are a copy editor. Cut this to [word count] words without losing [the specific detail that must survive]. Return the edit only.
• Rewrite this paragraph at a reading age of about 14, keeping every number and proper noun intact.
• Here is my draft and here is the brief. List every place the draft misses the brief, as a table of section, issue, and suggested fix.
• Rewrite this in the voice of the three sample paragraphs below. Match sentence length and vocabulary, not subject matter.
• Give me five alternative opening sentences for this piece, each taking a different angle: data, story, contrarian claim, question, and direct promise.
• Read this draft and list only the claims that would need a citation. Do not rewrite anything.
• Turn these rough notes into a structured outline with H2 and H3 headings. Do not write the body.
• Find every sentence in this draft longer than 30 words and rewrite just those.
• Rewrite this passive-heavy passage in active voice, keeping the same length.
• This piece is [word count] words and needs to be [target]. Show me what to cut, ranked by how little it costs the argument.
• Act as a sceptical reader of [publication]. List the three objections you would raise to this argument.
• Rewrite this for a reader who already knows [assumed knowledge], so we stop explaining the basics.
• Give me three title options under 60 characters and three under 40, all describing this piece accurately.
• Convert this article into a 200-word summary, then a 50-word summary, then a single sentence.
• Check this document for internal contradictions and list them with the line references.
• Rewrite these bullet points as flowing prose without adding any new information.
• Here is a transcript. Extract the five most quotable lines verbatim, with the speaker.
• Suggest where this draft needs a table, a chart, or an image, and say what each one should show.
• Rewrite this in British English and flag any terms that differ in meaning between British and American usage.
• Read this and tell me what the piece is actually arguing, in one sentence. If you cannot tell, say so.
• Give me a paragraph-by-paragraph critique focused only on whether each paragraph earns its place.
• Turn this long-form piece into a plain-text email newsletter of 300 words, keeping one link to the original.

Marketing and content

• Here is our product and our target buyer. Write five positioning statements, each leading with a different benefit.
• Take this feature list and rewrite each item as the outcome the buyer gets, in a two-column table.
• Draft a content brief for [keyword or question] including the search intent, the sections required, and the questions the piece must answer.
• We are launching [product]. Write the launch announcement in three lengths: a tweet, a 100-word post, and a 400-word blog intro.
• Read these three customer interview transcripts and list the exact phrases customers use to describe the problem.
• Turn this webinar transcript into six short posts, each with one takeaway and one supporting detail.
• Give me 20 content ideas for [audience] that are not already covered by these existing titles: [list].
• Write three versions of this landing page headline: one benefit-led, one problem-led, one specific and numeric.
• Audit this landing page copy against the four questions a buyer asks: what is it, who is it for, why now, what does it cost.
• Draft a five-email nurture sequence for someone who downloaded [asset] but has not booked a call.
• Here is our brand voice guide. Rewrite this copy to match it and note every change you made and why.
• Suggest ten hooks for a short video explaining [concept], each under 12 words.
• Turn this case study into a one-page PDF outline with a headline metric, the situation, the intervention, and the result.
• Write ad copy variants for [platform] within the character limits, five headlines and five descriptions.
• List the objections a [job title] would raise about buying [product], and draft one sentence answering each.
• Build a 90-day content calendar for [audience] with one pillar piece and three supporting pieces per month.
• Rewrite this press release so a journalist could lift the first paragraph unchanged.
• Given this analytics export, tell me which three posts are worth updating rather than replacing, and why.
• Draft a survey of eight questions that would tell us whether [assumption] is true, avoiding leading phrasing.
• Write the FAQ section for this page using only questions a buyer would genuinely ask before purchase.
• Compare these two versions of our messaging and tell me which is more specific, with evidence from the text.
• Take this long-form guide and propose how to repurpose it across email, video, and a slide deck without duplicating it word for word.

Two of those touch the same discipline we cover at length in ChatGPT for content creation, which goes into where drafting with a model helps and where it quietly costs you more editing time than it saves.

SEO and AI search

• Here is a keyword and the top ten results. Tell me what search intent each result serves and which intent is underserved.
• Group this keyword export into topic clusters and name each cluster by the question it answers.
• For [keyword], list the subtopics that appear in at least three of the top ten pages and the ones that appear in only one.
• Write a page title and meta description for this page, under 60 and 155 characters, both including [keyword] naturally.
• Read this page and suggest FAQ questions based only on gaps a reader would still have after finishing it.
• Given this list of published URLs, suggest internal links for a new page about [topic], with the anchor text.
• Turn this page into an answer-first structure: state the direct answer in the first 60 words, then support it.
• Draft FAQPage schema for these five questions and answers, as valid JSON-LD.
• List the entities a page about [topic] should mention to be considered comprehensive, and mark which ones my draft is missing.
• Here is my article. Extract the three sentences most likely to be quoted verbatim by an AI answer engine, and explain why.
• Compare my page against this competitor page and list what they cover that I do not, as a table.
• Rewrite these H2 headings so each one is a question a person would actually type or ask.
• Given this Search Console query export, find queries where we rank between 8 and 20 and the page does not directly answer the query.
• Suggest which of these 30 pages should be merged, which updated, and which left alone, with a one-line reason each.
• Write a paragraph that defines [term] in under 50 words, in a form a model could lift as a definition.
• List the buyer questions in [category] that a person would ask a chatbot rather than type into a search box.
• Read this product page and tell me what a model would struggle to extract about pricing, availability, and who it is for.
• Draft alt text for these images that describes the content rather than repeating the keyword.
• Given this topic, list the third-party surfaces where a recommendation would carry more weight than our own blog.
• Audit this article for claims stated without a number, and suggest what data would make each one citable.

That last cluster is where classic SEO and AI search start to diverge. Ahrefs' analysis of what correlates with AI visibility found the strongest relationship with third-party mentions and video rather than on-page work, which is why several of those prompts point away from your own pages. The mechanics are in how AI models choose which brands to recommend.

Sales and outreach

• Here is a prospect's website and their recent announcement. Draft a 90-word email referencing something specific, not generic praise.
• Rewrite this cold email so it makes one ask and removes every sentence about us.
• Given this call transcript, list the buying signals and the unresolved objections separately.
• Draft three follow-up emails for a deal that has gone quiet after [stage], each with a different reason to reply.
• Write a one-page briefing on this account: what they do, who the likely buyer is, and three questions worth asking.
• Turn our pricing page into a plain answer to the question what will this cost me, for a [company size] buyer.
• List the ten questions a technical evaluator would ask about [product] and draft honest answers, including where we are weak.
• Rewrite this proposal so the commercial terms are on the first page and the detail follows.
• Given this objection, draft a response that concedes the valid part before answering it.
• Draft a discovery call agenda for a 30-minute first call with [job title] at a [industry] company.
• Summarise this email thread into a status, a next step, and an owner.
• Write a LinkedIn connection note under 300 characters that references [specific detail] and asks nothing.
• Turn these three case studies into a one-paragraph proof point for a [industry] prospect.
• Given our win-loss notes, list the three reasons we lose most often and one testable change for each.
• Draft a renewal email for a customer whose usage has dropped, without implying they have done something wrong.
• Write the internal handover note from sales to onboarding for this closed deal.
• Rewrite this sequence so no two emails open with the same sentence structure.
• Given this list of 50 accounts and our ideal customer profile, rank them and explain the top five rankings.

Want to get recommended by AI?

Check your AI search visibility, then let the Honeyb agent write, fix, and earn what gets you recommended. Free to start.

Free AI visibility checker

Research and competitive analysis

• Compare these three competitors on pricing, positioning, and target buyer, using only what is on their public pages. Cite the page for each claim.
• Read this competitor's homepage and tell me who they are excluding, based on the language they use.
• Summarise this 40-page report into the five findings that would change a decision, with the page number for each.
• List what this study does not measure, and where its conclusions are being over-extended.
• Given these ten customer reviews, group the complaints by root cause rather than by wording.
• Build a table comparing [category] tools on the four criteria a buyer actually decides on.
• Here are two sources that disagree. Lay out the disagreement, the evidence each uses, and what would settle it.
• Read this pricing page and tell me exactly what a customer pays in year one, including anything conditional.
• Identify the assumptions in this business case that, if wrong, would break it.
• Given this market overview, list the segments that are described but never sized.
• Summarise this earnings call into what management said about [topic], quoting directly.
• Turn these interview notes into a themes table with the count of interviewees who raised each theme.
• List the questions this research answers, and separately the questions a reader would still have.
• Read these job adverts from a competitor and infer what they are building, flagging where you are speculating.
• Given this dataset description, tell me what analysis would be invalid because of how it was collected.
• Extract every number from this document into a table with the figure, what it measures, and the stated source.
• Draft a research plan to test [hypothesis] with a budget of [amount] and four weeks.
• Fact-check each claim in this passage and mark it supported, unsupported, or unverifiable from the text alone.

Data and spreadsheets

• Write a spreadsheet formula that [describe the calculation], and explain what each argument does.
• Here is my formula and the error it returns. Find the cause and give me the corrected version.
• Given these column headers and three sample rows, suggest the cleaning steps this dataset needs before analysis.
• Write a SQL query that answers [question] against these tables, and note any assumption you had to make about the joins.
• Explain what this query does in plain language, then list the edge cases where it would return the wrong result.
• I have [describe the data]. Which chart type answers [question] most honestly, and which would be misleading.
• Turn this pivot table into three sentences a non-analyst would understand.
• Given these results, tell me what could explain the change other than the thing we are hoping caused it.
• Suggest a validation rule set for this data entry sheet so bad values cannot be entered.
• Write a formula that flags rows where [condition], and returns blank otherwise.
• Here is a monthly series. Tell me whether the recent movement is outside normal variation, and show your working.
• Convert this wide table to long format and explain when each shape is preferable.
• Given this sample size and split, is the difference between these two groups meaningful, and what would I need to be confident.
• Draft the definition of each metric in this dashboard, precise enough that two people would calculate it identically.
• Review this analysis and list every place a percentage is quoted without its base.
• Write a step-by-step plan to reconcile these two reports that disagree by [amount].

Coding and technical

• Here is the error and the surrounding code. Give me the most likely cause first, then the next two, with how to test each.
• Review this function for correctness only. Ignore style. List issues with the line number and a failing input.
• Refactor this for readability without changing behaviour, and list every change you made.
• Write unit tests for this function covering the happy path, the boundaries, and the error cases.
• Explain what this code does line by line, then tell me what it does that a reader would not expect.
• This query is slow. Given the schema and the plan, tell me what to index and why.
• Convert this script from [language] to [language], keeping the same structure so I can diff them.
• Here is a stack trace. Tell me which frame is mine and which is library, and where to start.
• Write the docstring for this function, documenting the arguments, return value, and raised errors.
• Review this API design and list the decisions that will be painful to change later.
• Given these requirements, list the edge cases the specification does not address.
• Write a regular expression that matches [pattern], and give me three strings it should match and three it must not.
• Turn this bug report into a minimal reproduction case.
• Explain the trade-off between [approach A] and [approach B] for this specific workload.
• Review this migration for anything that would lock a large table or lose data.
• Write a shell command that [task], and explain the flags before I run it.
• Here is my test that fails intermittently. List the possible sources of nondeterminism.
• Draft the README for this project: what it is, how to run it, and the three things a new contributor gets wrong.
• Given this dependency list, flag anything unmaintained or duplicated in function.
• Write a post-incident summary from this timeline: impact, cause, resolution, and the two changes that would prevent recurrence.

Business strategy and operations

• Given these costs and this price, calculate the gross margin and tell me what breaks it at [volume].
• Write a one-page decision memo for [decision]: the options, the criteria, the recommendation, and what would change it.
• Here is our current pricing. List three alternative structures and who each one favours.
• Map this process step by step and mark every point where work waits on someone.
• Draft the standard operating procedure for [task], written so a new starter could follow it unaided.
• List the ways this plan fails, ranked by likelihood, then by cost if it happens.
• Given this budget, tell me which three line items carry the most risk of overrunning and why.
• Turn these quarterly goals into measurable outcomes with a named owner and a number.
• Write the agenda for a 60-minute meeting that must produce [specific decision].
• Summarise this contract into obligations, deadlines, and termination terms, flagging anything unusual.
• Given this org chart and these complaints, suggest where the reporting lines are causing the friction.
• Draft a build-versus-buy comparison for [capability] with the costs neither side usually mentions.
• Here is our churn data by cohort. Tell me what question I should be asking that I am not.
• Write the two-page brief a board member would need to understand [initiative] in five minutes.
• List the metrics we should stop reporting because nobody acts on them, based on these dashboards.
• Draft a risk register for this project with likelihood, impact, and the owner of each mitigation.
• Given these three vendor quotes, build a comparison table on total cost over three years.
• Write the pre-mortem: it is a year from now and this failed. What happened.

Customer support

• Here is the customer message and our policy. Draft a reply that is accurate, under 120 words, and does not blame the customer.
• Rewrite this support macro so it sounds like a person and still covers every required point.
• Given this angry message, draft a reply that acknowledges the specific problem before offering the fix.
• Turn these 20 tickets into a table of issue, frequency, and whether it is a bug, a gap in documentation, or a design problem.
• Draft the escalation summary for engineering: what the customer sees, what we know, and what we need.
• Write the help-centre article for [feature], structured so the answer is in the first paragraph.
• Here is a reply I drafted. Tell me where a customer could reasonably misread it.
• Draft a message telling a customer we cannot do what they asked, offering the nearest thing we can.
• Turn this changelog into a customer-facing note that explains why the change helps them.
• Given these recurring questions, propose which three should become in-product copy rather than support articles.
• Write an outage notification for [situation] with an honest status and no speculation about the cause.
• Draft the follow-up we send after a resolved ticket that does not ask for a review.
• Rewrite this refund policy in plain language without changing what it commits us to.
• Given this conversation, tell me at which message the customer decided we were not listening.

Hiring and management

• Rewrite this job advert to describe the work rather than the ideal person, and remove every unnecessary requirement.
• Given this role, draft six interview questions that test the actual work, not general reasoning.
• Write a take-home exercise for [role] that takes under two hours and reflects real tasks.
• Here are my interview notes on two candidates. Lay out the evidence for each against the criteria, without recommending.
• Draft a 30-60-90 day plan for a new [role], with what success looks like at each stage.
• Write feedback for [situation] that is specific, describes impact, and proposes one change.
• Turn these performance notes into a review that separates what happened from how I feel about it.
• Draft the agenda for a first one-to-one with a new report.
• Given this team's workload, suggest what to stop doing before we ask for another hire.
• Write a promotion case for [person] built on evidence rather than adjectives.
• Draft the message announcing [organisational change] that answers the questions people will actually have.
• List the questions a candidate should ask us about this role, and draft honest answers.
• Rewrite this rejection email so it is brief, kind, and does not offer false encouragement.
• Given these exit interview notes, identify the themes that management can act on.
• Write a scorecard for this role: four competencies, each with what strong and weak evidence looks like.
• Draft a delegation brief for [task] that includes the outcome, the constraints, and the decisions the person owns.

Learning and productivity

• Explain [concept] to me twice: once for a beginner, once for someone with a working knowledge of [related field].
• I understand [concept A] well. Explain [concept B] by mapping it onto what I already know, and note where the analogy breaks.
• Give me a four-week study plan for [subject] at five hours a week, with what to do each week.
• Ask me ten questions to find out where my understanding of [topic] is weakest. Do not explain anything yet.
• Here is my summary of [concept]. Correct only what is wrong, and say what is right.
• Turn this chapter into flashcards, question on one side, answer on the other.
• Explain the three most common misconceptions about [topic] and why each one is tempting.
• Give me the ten terms I need to know before reading this paper, defined in one sentence each.
• Here is my to-do list and my available hours. Tell me what will not get done and propose which to drop.
• Turn this vague goal into three concrete next actions I could start today.
• Read my calendar for the week and tell me what proportion of my time is committed before I have done any work.
• Given this decision I keep postponing, list what information I am actually waiting for and whether it will arrive.
• Summarise this meeting transcript into decisions made, actions with owners, and open questions.
• Write the email declining this request politely and without a long explanation.
• Given this project, break it into tasks no longer than half a day each.
• Ask me one question at a time to help me write my weekly plan. Wait for each answer.
• Explain what I would need to believe for [decision] to be the right one.
• Here is a topic I have to present on. Give me the five slides and the one number each must carry.

Brand and AI visibility

These are the prompts worth running about your own company rather than for it. Buyers are typing versions of these into ChatGPT before they ever reach your website, and the answers are shaping a shortlist you do not see. Run each one several times, in a fresh chat, and log what comes back.

• What are the best [category] tools for a [company size] [industry] company, and why.
• I need [outcome]. Which three products should I shortlist, and what are the trade-offs.
• What is [your brand] and who is it for.
• What do people dislike about [your brand].
• Compare [your brand] and [competitor] on price, features, and who each suits better.
• Is [your brand] a good choice for [specific use case]. Explain your reasoning.
• What does [your brand] cost, and what is included at each tier.
• What are the main alternatives to [competitor], ranked.
• List the sources you used to answer that, with links.
• What do you know about [your brand] that is more than a year old and might be out of date.
• Who are the credible companies in [category] that a cautious buyer would consider.
• What would make you not recommend [your brand].
• If I asked you this same question tomorrow, what might change about your answer.
• Summarise the reputation of [your brand] based on what you have seen, and say how confident you are.

Two things to expect when you run these. Answers vary run to run, for the reasons set out above, so a single run tells you very little. And a large share of what shapes them never names you at all: Semrush found that 62% of AI citations do not mention the brand, so the sources influencing your shortlist position are often pages about your category rather than pages about you. If the answers come back wrong or absent, why your brand is not showing up in ChatGPT covers the usual causes.

Doing this at scale, disclosed as ours

We build Honeyb, so treat this as disclosure rather than a neutral recommendation. The prompts in the section above are the manual version of what Honeyb automates: it runs buyer-intent prompts on a schedule across ChatGPT, Perplexity, Google AI Mode and AI Overviews, Gemini, Claude and Copilot, repeats them enough times to see past the run-to-run variance, and tracks how often you are mentioned, your share of voice against the competitors named alongside you, and the sentiment of the framing. Doing that by hand across seven engines and a few dozen prompts is a morning a week, which is exactly the sort of task that quietly stops happening. The arithmetic behind the numbers is in how to measure AI share of voice.

If you only take one prompt from this page, take the first one in the brand section, and run it about your own category. Then run it four more times. The spread between those five answers is the honest picture of where you stand, and it is usually wider than anyone expects. You can see the same thing without the copy-pasting with a free AI visibility check, which runs buyer prompts about your brand across the major engines and reports what they say back.