AI Prompt Generator

An AI prompt generator that builds, checks and shares your prompt

This AI prompt generator turns a few plain answers — who the AI should be, what it should do, what it needs to know and what the result should look like — into a structured prompt you can paste into ChatGPT, Claude, Gemini or any other chat assistant. Start from one of 19 templates across nine use cases, or from a blank form, and the prompt builds itself as you type.

A rule-based checklist reviews it as you go: a missing goal, vague verbs such as “help with”, no output format, or instructions that contradict each other, like “brief” next to “comprehensive”. Everything, including favourites and share links, stays in your browser.

Writing for one assistant? The ChatGPT, Claude and Gemini generators follow each vendor’s own advice, and the coding prompt generator handles code and error logs.

How to build a prompt

1. Pick a template under Start from a template, or “Blank — start from scratch”. Undo brings back what you had.
2. Fill in any variables under Fill in the variables. Type {{anything}} in a box to add your own.
3. Adjust The request, then Format and style, then Steps, rules and examples.
4. Fix anything the Prompt checklist marks Fix or Check.
5. Choose Plain text, Markdown or JSON and press Copy, or use Copy share link and Save to favourites.

AI Prompt Generator features

What goes into a good prompt

OpenAI, Anthropic and Google each publish prompting guides, and on the basics they agree: state exactly what you want, give the context the model cannot guess, show an example and name the format. The form maps onto that advice:

Sources, checked 14 October 2026: OpenAI’s “Prompt engineering” guide and Help Center best-practices article, Anthropic’s “Prompting best practices” and Google’s “Prompt design strategies” for the Gemini API.

How the checklist decides

The checklist is a fixed set of rules, not an AI, so the same prompt always gets the same verdict. Four key checks are scored:

• Goal — the task is empty (Fix) or under four words (Check).
• Vague wording — “help with”, “deal with”, “work on”, “look into”, “make it better”, “stuff”, “etc.” and similar.
• Output format — none chosen and no format word (list, table, JSON, email…) in the task or rules.
• Conflicts — short versus detailed, formal versus casual, the chosen format versus another named in the text, and word limits that cannot all be met.

Unscored tips point out a missing role, audience or examples, empty variables, JSON without a schema and rules that only say “don’t”. Rules can miss subtle problems, so treat them as a second pair of eyes.

Worked example

Load “Product description” and fill the variables with “the Aurora insulated steel bottle (750 ml)”, its features, “$29” and “commuters and gym-goers”. The plain-text prompt reads:

You are an e-commerce copywriter.

Task:
Write a product description for the Aurora insulated steel bottle (750 ml).

Context:
Key features: keeps drinks cold for 24 hours and hot for 12, leak-proof lid, fits most car cup holders
Price: $29

Audience:
commuters and gym-goers

Rules:
- Lead with the main benefit, not the feature list
- Use only the features given; do not invent specifications
- End with one sentence that invites the reader to buy

Output format:
Write in flowing prose paragraphs.
Tone: persuasive.
Length: about 120 words.

All four key checks pass; a tip suggests adding an example. The JSON view turns the same prompt into a system message (the role) and a user message (the rest, in Markdown).

Share links and the token estimate

Copy share link compresses the form with your browser’s built-in deflate and stores it after the # in the address. Browsers never send that part to a web server, so the prompt is not uploaded; whoever opens the link sees the same form.

The count under the preview divides characters by four, OpenAI’s rule of thumb for English. Real counts depend on the model and language — check them with the LLM token counter, and shorten long context with the text summarizer.