AI Prompt Generator

An AI prompt generator helps you write the instruction you give a chat model. A vague prompt gets a generic answer. A prompt that states the task, the background, the reader and the format gets something you can use.

This tool is a template builder, not an AI. It does not send your text to ChatGPT, Claude or Gemini. It arranges what you enter into a structure that works across models and points out what is missing.

You can output the prompt as Markdown headings or as XML tags. Anthropic’s prompt guidance recommends XML tags to structure prompts for Claude. For other models, Markdown headings are a safe default.

1. Pick the task type: This sets a default role, such as senior SEO strategist or data analyst. Use the role override if you know the exact role you want.
2. Describe the task: State what you want produced, for what purpose and how much. A line such as “write a 600-word intro for a post on X” beats “write about X”.
3. Add context: Paste facts, data, examples or the text to work from. Context is what separates a specific answer from a generic one.
4. Set audience, tone, format and rules: Say who will read the answer, the format and length you want, and any rules such as words to avoid or sources required.
5. Choose a style and extras: Pick Markdown or XML style. Tick the options that let the AI ask questions first, check its own answer, or avoid filler.
6. Copy the prompt: Copy the full prompt or the short version into your chat model. If the completeness score is low, fix the gaps listed under it first.

An SEO writes meta descriptions for a physiotherapy clinic in Pune and sets format, length and rules.

Results shown are from the live tool, which also returns a short one-paragraph version and a completeness score (100 for this input).

• Turns half-formed requests into prompts with the parts a model needs, so you edit less after the first answer.
• Gives you a repeatable structure. Keep a prompt that works and swap only the task and the context next time.
• Shows a completeness score, so you know the weakest part before you run the prompt.
• Suits routine SEO work such as meta descriptions, content briefs, outreach emails and keyword grouping, where the same fields come up every time.

• Scores completeness and names the gap to fix, instead of only reformatting your text.
• Offers XML-tag output for Claude as well as Markdown headings.
• Builds a short one-paragraph version for chat apps where a long prompt feels heavy.
• Has switches for asking clarifying questions first, self-checking and banning filler.
• Needs no account, API key or paid plan, because nothing is sent to a model.

• It does not run the prompt. The answer still depends on the model you use and on the facts you supply.
• The completeness score is a checklist, not a prediction of answer quality. It checks that fields are filled and long enough, not that they are right.
• Role presets are generic. Replace them with the real role when you know it.
• No prompt guarantees accurate output. Check facts, numbers and code in anything a model returns.

The tool builds the prompt in your browser and does not store your input. Do not paste passwords, client data or other confidential material into a prompt you plan to send to a third-party model. That model may keep what you send.