Assistant Prompt Library
Best ChatGPT Prompts: Curated Examples for Real Tasks
ChatGPT is the most widely used AI assistant for everyday work. This collection covers what actually works well — organized around ChatGPT's real strengths: instruction-following, structured outputs, iterative refinement, and versatile task handling across almost any domain.
What makes ChatGPT prompting different
ChatGPT responds strongly to explicit role assignments, structured constraints, and clear output format requests. The most consistent improvement in ChatGPT output comes from three things: telling it what role to take ("Act as a..."), specifying the exact output format you need (table, list, paragraph, JSON), and giving it the context it needs to tailor the response to your situation rather than producing a generic answer.
ChatGPT also responds well to iterative prompting — the first response rarely needs to be final. Follow-up instructions like "shorten this by 40%," "rewrite the opening with a stronger hook," or "give me three alternative versions" consistently produce better output than trying to get everything right in one prompt.
Who these prompts are for
Writers, marketers, business owners, developers, researchers, educators, and anyone who uses ChatGPT regularly and wants to get more consistent, useful output by improving how they structure their prompts.
Best use cases for ChatGPT
- Writing first drafts and rewriting existing copy
- Structured analysis: pros/cons, option comparisons, frameworks
- Brainstorming: generating options, variants, and alternatives quickly
- Coding: debugging, explaining code, writing functions
- Summarization: compressing long documents or transcripts
- Education: concept explanation at different levels
Prompt examples
Role-based writing (core pattern)
Act as a [specific role — e.g. conversion copywriter / technical writer / senior editor]. Your task: [specific task]. Audience: [describe]. Output format: [format and length]. Tone: [tone]. Constraint: [one specific thing to avoid or include].
This is the single most reliable ChatGPT prompt pattern. The role sets the expertise frame; the constraint narrows the output.
Structured options
Act as a [role]. Give me [number] distinct options for [decision or problem]. For each: a short name, 2-sentence description, main advantage, and main tradeoff. Make the options meaningfully different — not variations of the same approach. Context: [describe your situation].
Add 'and then recommend one with your reasoning' to move from a list to a recommendation.
Rewrite and improve
Act as a senior editor. Rewrite the following [email / paragraph / headline / pitch] to be clearer, more direct, and easier to act on. Keep the core message. Improve the opening, cut filler, and strengthen the ending. Explain the 3 most important changes you made. Original: [paste].
Iterative refinement
I am working on [a blog post / proposal / email / landing page]. Here is my current draft: [paste]. Do not rewrite everything. Identify the single most important improvement to make, explain why it matters, and show only the revised version of that section.
ChatGPT handles targeted edits much better than full rewrites when you give it a specific section to fix.
Summarize with structured output
Summarize the following [article / report / transcript / meeting notes]. Output format: 1) Main argument or conclusion (2 sentences), 2) 3 supporting points (1 sentence each), 3) One significant limitation or caveat, 4) One actionable implication for [my role or decision]. Source: [paste].
Explain at multiple levels
Act as an expert teacher. Explain [concept] at three levels: beginner (no prior knowledge), intermediate (familiar with [related concept]), and advanced (can handle technical detail). Use a concrete analogy for the beginner level. Keep each explanation under 100 words.
Devil's advocate analysis
Act as a rigorous skeptic. I am planning to [describe plan or decision]. Challenge this plan: identify the 3 most likely failure modes, one assumption I may be making that is weaker than it looks, and one external factor I may have underestimated. Then suggest one specific adjustment that would make the plan more robust.
This prompt produces the most value when you disagree with the output — it surfaces things you had not considered.
Generate and filter
Act as a [role]. Generate 20 [ideas / headlines / angles / options] for [topic]. Then review your own list and identify the 5 strongest, explaining briefly why each one is stronger than the others. Context: [describe].
The self-review step significantly improves quality over just taking the top items from a list.
System prompt simulation
You are [role and persona description]. Your goal is [goal]. You always [key behavior 1]. You never [key behavior 2]. When asked about [topic], you [specific behavior]. Speak in [tone]. Now: [first task].
Useful for setting up a ChatGPT custom GPT persona or running a consistent-persona conversation across multiple turns.
Code with explanation
Act as a [language] developer. Write a function that [describe task]. Input: [describe]. Output: [describe]. Edge cases to handle: [list]. After the code, explain: what the function does in plain language, why you made each significant design decision, and one potential limitation to be aware of.
Common mistakes with ChatGPT prompts
- No role assignment: Skipping "Act as a..." produces noticeably more generic output. The role primes the response style and expertise level.
- No format instruction: Without a format request, ChatGPT defaults to flowing prose. Specify table, numbered list, bullet points, or JSON explicitly.
- Accepting first output as final: ChatGPT is designed for iteration. Follow up with targeted refinement requests instead of restarting from scratch.
- Trusting facts without verification: ChatGPT can produce confident-sounding but incorrect statistics, citations, and specific facts. Verify any factual claim before publishing or acting on it.
How to customize these prompts
Every bracketed field is a required input. Replace them with your specific role, audience, goal, tone, and constraints. The more concrete your inputs, the more specific and useful the output. If a prompt produces something generic, the fix is almost always in adding more specific context — not changing the prompt structure.
Related resources
- ChatGPT Prompt Framework
- ChatGPT Prompt Templates
- ChatGPT Model Guide
- ChatGPT Prompt Library
- Best ChatGPT Prompts for Business
- ChatGPT Prompt Generator
Use the ChatGPT Prompt Generator to create role-based, constraint-driven prompts for any task.
Open ChatGPT Prompt Generator