Stop Guessing: Find the Best AI Prompt for Any Job
Find the prompt by the job you need done
A prompt library is only useful if you can reach the right one in under a minute. Every category below is built around a real task, with a prompt you can paste, the inputs it needs, and a note on where it tends to fail.
Role: Senior B2B copywriter
Context: [product], [buyer], [objection]
Task: Write 3 subject lines
Constraints: Under 45 characters. No hype words.
Format: Table with a one-line rationale
Before you answer, list what you assumed.
Context: [product], [buyer], [objection]
Task: Write 3 subject lines
Constraints: Under 45 characters. No hype words.
Format: Table with a one-line rationale
Before you answer, list what you assumed.
The anatomy behind most prompts on this site. The last line is the one people skip.
Not sure where to start?
Pick the situation that sounds most like yours. Each path leads to a category, not a single prompt, so you can browse from there.
I write for a living
Start with Copywriting, then move to Marketing once you need campaigns instead of single pieces.
I teach or train people
Go to Prompts for Teachers. Read the policy notes before you build a lesson around AI output.
I want to understand the risks
Read AI Risk and Safety and AI Law and Copyright before you ship anything commercial.
All categories
Grouped by outcome. Each row shows what the category covers, a sample of its style, and the failure mode we see most often.
Copywriting
Ads, emails, landing pages, product copyPrompts that produce copy a person would sign off on: headlines with a point of view, emails with one clear ask, product pages that answer the real objection. Built to help you write faster without sounding like everyone else.
Rewrite this headline 5 ways. Each must name the buyer’s problem, not our feature. Flag any claim we cannot prove.
Where it breaks: models default to confident, generic superlatives. Ban them in the constraints line.
Browse copywriting promptsMarketing
Campaigns, positioning, answer-engine visibility, analyticsPositioning, audience research, campaign planning and reporting. This category favors prompts that make the model show its assumptions, because a plausible marketing plan built on a wrong assumption is the most expensive kind.
Act as a skeptical CMO. Review this campaign brief and list the three assumptions most likely to be wrong, with a cheap test for each.
Where it breaks: invented benchmarks. Never accept a statistic from a model without a source you can open.
Browse marketing promptsTeachers
Lesson planning, feedback, differentiation, AI policyPrompts for lesson plans, rubrics, feedback drafts and differentiated materials, paired with evidence on how teachers are actually using AI and what school policies now say. Time saved is only real if the output is checked.
Draft a 45-minute lesson on [topic] for [grade]. Include one misconception students commonly hold and a two-question check for it.
Where it breaks: age-inappropriate difficulty and confident factual errors. Review before it reaches students.
Browse teacher promptsResearch and analysis
Summaries, comparisons, source checking, synthesisPrompts for reading faster and thinking more carefully: structured summaries, side-by-side comparisons, and critique passes that look for what a source leaves out.
Summarize this document in 8 bullets. Then list what it does not say that a reader might assume it does.
Where it breaks: fabricated citations. Ask for quotes with locations, then verify them yourself.
Browse research promptsSEO and content strategy
Briefs, outlines, audits, refreshesPrompts for content briefs, outline critiques, update audits and refresh plans. The focus is on usefulness to the reader first, since search systems increasingly reward pages that add something a summary cannot.
Audit this article against the query “[keyword]”. List what a reader would still need to search for after reading it.
Where it breaks: keyword stuffing and filler sections. Cut anything that does not answer a question.
Browse SEO promptsFreelancing and business
Proposals, client emails, scoping, offersPrompts for the unglamorous half of independent work: proposals, scope notes, polite pushback, follow-ups and pricing conversations. Written to keep you in control of the message, not to outsource your judgment.
Turn these call notes into a one-page scope. Mark anything ambiguous as a question for the client instead of guessing.
Where it breaks: over-promising. Models happily commit you to deadlines you never agreed to.
Browse business promptsAI risk and safety
Evidence reviews, hype checks, risk analysisAnalysis rather than templates: what current research, surveys and safety reports say about where AI is useful, where it is risky, and how to tell a measured claim from a marketing one. Includes prompts for stress-testing AI output before you rely on it.
List the three strongest arguments against your own answer, then say which one you find most convincing and why.
Where it breaks: models agree with you. Ask for the opposing case explicitly.
Read the risk analysisAI law and copyright
Authorship, ownership, liability, disclosurePlain-language coverage of who owns AI-assisted work, what human authorship requires, and who is responsible when an output causes harm. Informational only, and not legal advice.
Explain this licensing clause in plain English. List the questions I should put to a lawyer before signing.
Where it breaks: rules vary by country and change fast. Check the date on any legal summary, ours included.
Read the legal guidesWhat separates a good prompt from a lucky one
Most weak prompts are not badly written. They are missing information the model cannot guess. Every prompt in the library is checked against the same short list.
- Role
- Who the model should act as, which sets vocabulary and level of scrutiny.
- Context
- The facts only you know: audience, product, constraints, what has been tried.
- Task
- One job, stated as a verb. Two jobs in one prompt usually means two mediocre results.
- Constraints
- Length, tone, banned words, what to leave out.
- Format
- The shape of the answer, so you can use it without reformatting.
- Self-check
- A final instruction to list assumptions or flag uncertainty. It costs one line and catches the most errors.
Weak
Write a product description for my coffee grinder.
Stronger
You are a product copywriter for a home-coffee retailer. Write a 90-word description of a burr grinder for buyers who currently use pre-ground coffee. Lead with the taste difference, not the specs. No “perfect” or “ultimate”. End by listing any claim you could not support from the details I gave.
How we keep the library honest
- Prompts are written for a job. If we cannot name the task a prompt solves, it does not get published.
- Failure notes are part of the entry. A prompt that works 7 times in 10 is still useful if you know which 3 to watch for.
- Facts link to sources. Where an article cites research, surveys or rulings, it names the source and says how certain the data is.
- Pages are dated and revised. Models change quickly. When a prompt stops working or a claim goes stale, the page is updated, not left to age.
- You stay responsible for the output. Prompts speed up drafting. They do not replace fact-checking, originality checks or professional advice.
Questions people ask before they start
Which prompt category should I start with?
Start with the category that matches a task you already do every week. A prompt you can test against your own work teaches you more in ten minutes than a long list of generic ideas.
Do these prompts work in ChatGPT, Claude and Gemini?
Yes. They are written in plain language with no tool-specific syntax. Models differ in tone and in how closely they follow format instructions, so expect to adjust the wording slightly the first time you use one.
How are prompts here different from a generic prompt list?
Each prompt is tied to a specific job, shows the inputs it needs, and includes a note on its known weak spots. Where an article makes a factual claim, it links to the source.
Can I use the prompts for client or commercial work?
You can use the prompt text freely. What you do with the output is your responsibility: check facts, check originality, and check the rules of your industry and your AI tool’s terms. The AI law category explains where the legal questions sit.
How often is the library updated?
Articles carry a visible updated date. When a model change, a new study or a legal ruling makes a page out of date, the page is revised.