AI Prompt Generator

Generate ready-to-use AI prompts for ChatGPT, Gemini and Claude. Writing, coding, creative and analytical prompts in one click.

What this tool does

This is a prompt builder, not an artificial intelligence. It hands you ready-to-use prompt templates for writing, coding, analysis, learning and creative work, which you then adapt and paste into whichever assistant you already use.

Nothing is generated by a model and nothing leaves your browser: the templates are stored on the page, the categories are simple lists, and there is no API call behind the buttons. That is deliberate - it means you can use the tool with any assistant, including ones running locally, and it means no prompt you write here is sent anywhere.

How to use it

Pick a category that matches the job you are doing.
Generate a prompt, or generate several and compare them.
Copy the one closest to your need and replace the bracketed parts with your own details.
Paste it into your assistant, then iterate: the first answer is a draft, not a verdict.

The anatomy of a prompt that works

Useful prompts tend to contain the same five parts, whether they are one line or ten:

Part	What it answers	Weak	Stronger
Task	What do you want done?	Help me with my resume	Rewrite the three bullet points below so each starts with a strong verb
Context	What should it know?	(nothing)	I am applying for a junior data analyst role; the reader skims for 20 seconds
Constraints	What are the limits?	Make it good	Maximum 20 words per bullet, no jargon, no invented metrics
Format	What shape should the answer take?	(nothing)	Return a markdown list of exactly three bullets, then one sentence of explanation
Examples	What does good look like?	(nothing)	Here is one bullet I already like: "Automated weekly reporting, cutting manual work by 6 hours"

The format row is the one people skip most often, and it is usually the cheapest to add. Asking for a specific shape - a table, a list of three, JSON with named fields - removes a whole class of answers you would otherwise have to throw away.

Prompt patterns worth knowing

Zero-shot versus few-shot. If you show one or two examples of what you want, the model copies their style and structure far more reliably than if you describe it.
Delimiters. Put pasted material between clear markers and say what the markers mean. It stops your instructions being read as part of the text you pasted.
Role framing. Telling the model who it is answering as changes vocabulary and depth, not facts, so treat it as a style control rather than a source of expertise.
Decomposition. Long jobs go better as separate prompts: outline first, then expand one section at a time, so mistakes stay local instead of spreading.
Output schema. When the result is going into a program, ask for strict JSON with the field names you need, then validate it.

Why vague prompts produce generic answers

An assistant answering a vague request has nothing to work with except what is most common in its training data, so it produces the average answer: the standard cover letter, the usual five tips, the generic lesson plan. Specificity is not politeness - it is the input that removes the average from the answer.

The most useful sentences in a prompt are often the negative ones. "Do not use the phrase leverage." "Do not invent statistics." "Do not add an introduction." Constraints are cheap to write and they cut revision time in half.

What prompts cannot do for you

They cannot supply facts you have not given. A model that does not know your numbers will produce plausible-looking ones. Treat every figure, citation and quotation in a generated answer as unverified until you check it.
They cannot replace domain judgement. A prompt can make a draft faster; it cannot make the decision about whether the draft is correct.
They are not confidential. Anything you paste into a hosted assistant leaves your machine. Do not paste client data, personal data or credentials into one.

A habit that improves every prompt

Keep the prompts that worked. When an answer is good, save the prompt next to it with a note about what you changed - the constraint you added, the example you included - and reuse it as a template. Most of the gain from prompting comes from a small library of prompts that fit your actual work, not from finding a cleverer phrase once.

Related tools: the JSON formatter is useful when you ask an assistant for structured output and want to check it parses, the diff checker helps when you are comparing two versions of a generated draft, and the word counter settles arguments about length limits.

FAQ

Does this tool use an AI model?

No. It is a library of prompt templates stored in the page. No model is called, no API key is needed, and nothing you type or generate is sent anywhere.

Which assistants do the prompts work with?

Any of them. The templates are plain text instructions, so they work in chat assistants, coding assistants and locally hosted models alike - the results differ, but the prompt does not need rewriting.

Can I use the generated prompts commercially?

Yes. They are generic instructions, not creative works, so adapt them freely. Check your assistant provider's terms for rules about the output rather than the prompt.

How do I stop getting generic answers?

Add the two parts people usually omit: constraints and format. Say what to avoid, set a length, and ask for a specific shape such as three bullets or JSON with named fields.