AI Prompt Examples: 30+ Templates That Work

AI prompt examples are copy-ready instructions you paste into ChatGPT, Claude, or Gemini and adapt to your case. The best examples give the model a role, clear context, a precise task, and an output format. This guide collects 30+ ready-to-use examples grouped by use case — and shows you how to turn each one into your own reusable template.

A good example is worth more than any abstract theory. You don't need to know what "chain-of-thought" means to use a prompt that thinks step by step. That's exactly why this article is a pure example collection: real, tested prompts you can select, copy, and use right away. Theory and background are linked at the relevant points if you want to go deeper.

Three notes up front. First, these examples work across models — whether ChatGPT, Claude, or Gemini, the principles stay the same. Second, every placeholder is written in curly braces like {topic}; replace them with your specifics. Third, it pays to permanently save the examples you use often, so you never retype them. At the end we'll show you exactly how.

What makes a good prompt example?

A good prompt example is concrete, copy-ready, and contains clearly visible spots you adapt. It combines four building blocks: a role, context, a task, and a format. This anatomy is what separates a useful example from a vague question.

Look at the difference. A vague question:

Prompt

Write something about time management.

A strong example with all four blocks:

Prompt

You are an experienced productivity coach. I'm a freelancer with too many parallel projects. Write three concrete time-management methods I can test this week. Deliver the result as a numbered list, each with one sentence of explanation and a first step.

The second example reliably produces usable answers because it leaves nothing to chance. You'll recognize the same blocks in every good example:

If you want the depth behind this anatomy, our guide to [writing effective AI prompts](/magazin/write-effective-ai-prompts) explains each block in detail. For a quick start, the examples below are enough — you don't need the theory to use them.

Prompt examples for writing

For writing, AI shines as a sparring partner for structure, revision, and variations. These examples cover the most common writing tasks.

Create an outline:

Prompt

You are an experienced editor. Create an outline for an article about {topic} for {audience}. Five main points, each with one key message in a single sentence. Logical flow from problem to solution.

Revise a paragraph:

Prompt

Revise this paragraph: clearer, more active, no filler words. Keep my tone. Show the changes and justify each in five words max. Text: {paste paragraph here}.

Continue in my style:

Prompt

Here are two paragraphs in my style: {sample texts}. Write the next paragraph about {topic} in exactly this style — same sentence structure, same voice.

Make a text more concrete:

Prompt

Make this text more concrete. Replace every general statement with a precise example, a number, or a scene. Keep the length the same. Text: {paste text}.

Generate headlines:

Prompt

Write ten headlines for an article about {topic}. Five curiosity-driven, five benefit-driven. Max 60 characters, no clickbait.

The common thread: you always give the model style, audience, and purpose. For a much larger collection specifically for ChatGPT, see our overview of the [best ChatGPT prompts](/magazin/best-chatgpt-prompts).

Prompt examples for coding

For developers, AI is strongest at explaining, testing, debugging, and refactoring. The key: always supply full context — code, error message, and expected behavior.

Debug an error:

Prompt

Here is my code and the error message. Explain the most likely cause and propose a fix. Justify why the error occurs. Code: {code} — Error: {error message}.

Write tests:

Prompt

Write unit tests for this function. Cover edge cases and failure cases. Framework: {Jest/Pytest/...}. Function: {code}.

Explain code:

Prompt

Explain this function line by line as if I were a junior developer. Point out possible pitfalls. Code: {code}.

Refactor:

Prompt

Refactor this code for readability without changing behavior. Explain each change in one sentence. Code: {code}.

Build a regex:

Prompt

Write a regular expression that matches {describe pattern}. Explain each part and give three example matches and three non-matches.

Always treat generated code as a draft that needs review and tests — never as a finished solution. When debugging especially, state the expected behavior explicitly so the model doesn't guess past the actual problem.

Prompt examples for marketing

In marketing, AI delivers best when you ask for variations instead of a single version. Always request several options to choose from and test.

Subject lines:

Prompt

You are an email marketing expert. Write ten subject lines for a campaign about {product}. Max 50 characters, five curiosity-driven, five benefit-driven. No clickbait.

LinkedIn post:

Prompt

Write a LinkedIn post about {topic} for {audience}. Hook in the first line, one concrete example in the middle, a question at the end. Tone: personal, no marketing speak. 150 words.

Value proposition:

Prompt

Write three variations of a value proposition for {offer}. One sentence each, focused on the outcome for the customer, not on features.

Ad copy:

Prompt

Write three Google Ads variations for {product}. Each with one headline (max 30 characters) and one description (max 90 characters). Focus: {main benefit}.

Product description:

Prompt

Write a product description for {product} in three sentences. Neutral tone, ending with one concrete benefit promise. Audience: {audience}.

AI marketing copy is only as good as the brief. Feed it an audience, a tone, and a real customer benefit, and you get on-brand copy — type only "Write me an ad" and you get interchangeable filler.

Prompt examples for research and analysis

For research and analysis, AI is strongest as a structurer and sparring partner. Important: for current facts you need web search or your own sources, since the model can invent numbers.

SWOT analysis:

Prompt

Create a SWOT analysis for {initiative}. Three points per quadrant, concrete and verifiable. Deliver the result as a table.

Weigh options:

Prompt

Weigh {option A} against {option B}. Score each criterion from 1 to 5, present it in a table, and justify the recommendation.

Check assumptions:

Prompt

What unspoken assumptions are baked into this plan? List the three riskiest and say how I could verify each. Plan: {plan}.

Devil's advocate:

Prompt

Argue against my conclusion. What are the strongest counterarguments, and what data would disprove my thesis? Thesis: {thesis}.

Structure data:

Prompt

Here is raw data: {data}. Structure it into a table, name three anomalies, and one hypothesis.

A strong meta-prompt for any research: "Before you answer, tell me what information you're missing to deliver a well-founded analysis." This stops the model from filling gaps with plausible inventions.

Prompt examples for business and productivity

In daily work, AI saves the most time on recurring routine tasks: emails, meetings, planning, and decisions.

Draft an email:

Prompt

You are my assistant. Turn these bullet points into a polite, clear email to {recipient}: {bullet points}. Tone: {tone}. Max 120 words. End with one concrete next step.

Summarize a meeting:

Prompt

Summarize these notes, split into decisions, open points, and action items with owners. Notes: {notes}.

Plan the week:

Prompt

Here are my open tasks and deadlines: {tasks}. Prioritize by urgency and effort, propose a realistic daily split for this week, and flag what I could delegate or drop.

Difficult feedback:

Prompt

I need to give {person} critical feedback about {topic}. Write three conversation openers using the SBI model (Situation, Behavior, Impact), respectful but clear.

Structure a decision:

Prompt

Help me decide between {A} and {B}. First ask me the five questions I need to answer before you give a recommendation.

For a broader collection of such templates plus a system to save them, see the guide to the [best ChatGPT prompts](/magazin/best-chatgpt-prompts).

How do you adapt an example to your needs?

You adapt an example by replacing the placeholders in curly braces with your specifics and adding a constraint where needed. Those placeholders are exactly what turn a single example into a reusable template.

Take this scaffold as a universal template:

Prompt

You are an experienced {role}. My situation: {context}. Your task: {goal}. Deliver the result as {format}. If you're missing information, ask me the three most important questions first.

Three steps turn any example into your own template:

1. Replace placeholders — fill {role}, {context}, {goal}, and {format} with your details. 2. Add a constraint — length, tone, or banned words where relevant ("max 100 words", "no jargon"). 3. Iterate — treat the first answer as a draft and sharpen it: "Make the second paragraph more concrete", "Give me three shorter variations".

Keep your placeholders consistent — always in the same brace format — so you see at a glance what to adjust. If you want to start with no typing at all, have a fitting prompt generated directly by the [AI prompt generator](/tools/prompt-generator) and use it as a starting point.

Where should you store your prompts?

You best store good prompts in a dedicated prompt library rather than the chat history — with a clear title, tags by use case, and placeholders for the parts that change. That turns a one-off hit into a template you find again in seconds.

The chat history is unsuited as storage: sorted chronologically rather than thematically, barely searchable, and locked to a single tool. As soon as an example from this article has worked twice for you, promote it into your library. Three principles hold even at hundreds of prompts:

• Central, not scattered — all prompts in one place, not across notes, chats, and documents at once.
• Tag by use case — "Marketing", "Code", "Email" instead of by tool, since models change but tasks stay.
• Version — when you improve a prompt, keep the old version so you can roll back a regression.

For how to build such a system step by step — with folders, tags, and naming conventions — see our guide to [building a personal prompt library](/magazin/build-personal-prompt-library). That's exactly what Prompt2Love is built for: a central library where your best prompts are saved, tagged, and shared across your team.

Frequently Asked Questions

What is an example of a good AI prompt?

A good example is: "You are an expert copywriter. Write three subject lines for a B2B software, max 50 characters, neutral tone." It contains a role, task, format, and constraint — and therefore reliably produces usable answers instead of vague generalities.

How many examples should I give in a prompt?

For most tasks, one to three examples (few-shot) are enough. They show the model the desired pattern more reliably than any abstract description. More than three rarely adds value and needlessly lengthens the prompt.

Do these examples also work with Claude and Gemini?

Yes. The core principles — role, context, task, format — apply to every language model. Only fine details like tone strength or format specifications differ. Tag your prompts by task, not by tool: the task stays, the model is interchangeable.

Should I prompt in English or another language?

Both work excellently. Prompt in the language you need the result in. Modern models handle most major languages nearly on par with English; only for very specialized terminology can English be marginally more precise.