The AI prompt library

The best prompt for every task.

Find the right prompt. Get better results from AI.

66 prompts12 categories5 AI tools coveredFree to copy, no account

Explore by category

Twelve areas of professional work, each with a hub that explains which prompt to reach for.

All categories →

Editor’s picks: the broadest, most complete prompt in each category. See how we choose them.

Browse all 66 prompts →

Explore by profession

Curated sets that cut across categories, with a note on when to use each prompt.

All collections →

Explore by AI model

The prompts here are written to be model-independent: plain structure, explicit inputs, a stated output format. Each prompt page lists the tools it is expected to work with and any adjustment worth making. A prompt is only labelled Tested for a model when a test record exists.

How model compatibility works →

ChatGPT prompts

How to prompt ChatGPT: clear instructions, the right context, and standing preferences set once instead of repeated.

66
Claude prompts

How to prompt Claude: explain the task as you would to a capable colleague, structure long inputs, and show examples.

66
Gemini prompts

How to prompt Gemini: specific instructions, stated constraints, a few examples, and the question after the context.

66
Perplexity prompts

How to prompt Perplexity: ask a precise research question, say which sources count, and check the citations it returns.

3
Cursor prompts

Cursor is an AI code editor, not a model. To prompt it well: reference the right files, set project rules, review diffs.

11

Build your own with the Prompt Builder

Not sure where to start? Answer eight short questions and get a structured prompt you can copy and edit. Runs in your browser.

Open the Prompt Builder

Recently reviewed prompts

Taken from the editorial review log. Every entry is a real record with a date and a review type.

PromptCategoryReviewDateTesting status
Core Web Vitals Diagnosis PromptWeb DevelopmentStructural reviewUntested
Technical Specification PromptWeb DevelopmentStructural reviewUntested
React Component PromptWeb DevelopmentStructural reviewUntested
Next.js App Router Feature PromptWeb DevelopmentStructural reviewUntested
Relational Database Schema PromptWeb DevelopmentStructural reviewUntested
Responsive CSS Layout PromptWeb DevelopmentStructural reviewUntested

Learn to write better prompts

Guides on the techniques behind the library, grounded in the model vendors’ own documentation.

All guides →

How prompts are developed, tested and maintained

We would rather label a prompt honestly than call everything “tested”.

Read the editorial standards →
  1. Developed

    Each prompt starts from a real task and the inputs a practitioner would ask for, with explicit constraints against invented facts and padding.

  2. Reviewed

    Every prompt is checked for structure: variables resolve, the output format is stated, limitations are listed and sources are real.

  3. Tested

    A prompt is labelled Tested only when a test record exists: model and version, inputs, actual output and notes against a seven-point rubric. No invented scores.

  4. Maintained

    Prompts are versioned. When a model changes or feedback points to a problem, the prompt is flagged for retesting and the label changes with it.