AI Prompt Generator

Turn a rough idea into a structured prompt for ChatGPT, Claude, Gemini or Midjourney. Each prompt is laid out the way that AI's own documentation recommends. Free, instant, no sign-up.

The prompt updates as you type. Paste it into your AI, then answer any questions it asks.

What this AI prompt generator does

Most weak AI answers come from weak prompts, not weak models. A one-line request like "write a blog post about Japan" leaves the AI to guess who it is writing for, how long it should be, what to avoid and what the finished result should look like, so it falls back to the most generic answer. This tool asks you for those missing pieces in plain language and assembles them into a complete prompt you can paste into any AI.

It runs entirely in your browser. Nothing you type is sent to us or to any AI until you paste it yourself, there is no sign-up and no daily limit.

Why the prompt looks different for ChatGPT, Claude and Gemini

Most prompt generators give every model the same block of text. This one changes the layout to match what each company's own documentation recommends, because the models are trained to read structure in slightly different ways:

  • Claude gets its sections wrapped in XML-style tags such as <context>, <task> and <requirements>. Anthropic's prompting guide recommends this directly: tags help Claude separate your instructions from your background material and reduce misreading.
  • ChatGPT gets Markdown headings (# Role, # Task, # Output format). OpenAI's prompt engineering guide suggests combining Markdown and XML to mark the boundaries of each part, and lays prompts out in named sections such as identity, instructions and examples.
  • Gemini also gets clearly headed sections. Google's Gemini prompt design guide advises using clear delimiters, XML-style tags or Markdown headings, and putting the role and output-format requirements first.
  • Other chatbots get a plain-text version with a short opening paragraph and a requirements list, which works in almost any tool.

The content is the same in every version. Only the packaging changes, and for longer prompts with lots of context that packaging makes a noticeable difference to how closely the AI follows your instructions.

The five parts of a strong AI prompt

The checklist under your prompt tracks five elements. You don't need all of them every time, but each one you add removes a guess the AI would otherwise make:

  1. Role - who the AI should act as. "A senior copywriter" writes differently from "a patient teacher". The generator picks a sensible role for each task type.
  2. Task - the specific thing you want done, stated as an instruction. This is the only part you must fill in.
  3. Context - who it is for and the background the AI can't know: your product, your situation, facts that must be included.
  4. Constraints - tone, length and anything to avoid. Constraints are what stop the answer drifting into generic filler.
  5. Output format - what the finished answer should look like: a table, numbered steps, JSON or a code block. Stating it saves a round of "can you put that in a table?".

Each task type also adds two or three quality rules of its own. An email prompt asks for the main point in the first two sentences and a specific subject line; a coding prompt asks the AI to state its assumptions and cover edge cases; a research prompt asks it to say when it isn't sure instead of guessing.

The "ask me questions first" option

It is switched on by default because it is the single most useful line you can add to a prompt. It tells the AI to check what it doesn't know before it starts, so instead of confidently filling gaps with assumptions it asks two or three quick questions. Answer them and the first draft is usually close to what you wanted. Turn it off when you want an immediate answer.

Image prompts: Midjourney vs ChatGPT

Image models read prompts differently from chat models, so the generator switches layout when you choose an image AI:

  • Midjourney gets a compact, comma-separated description followed by parameters. Midjourney's documentation says parameters always go at the end of the prompt, so the shape is added as --ar 16:9 and anything to exclude as --no at the very end. No model version is forced, so your default version is used.
  • ChatGPT images (DALL-E) get short, complete sentences, because ChatGPT interprets your request conversationally rather than as a keyword list. Exclusions are written out as "Do not include...". For logos it also asks for any text to be kept minimal and spelled exactly, since image models still misspell words.

How to get the best results

  1. Pick the AI first. The layout, the "Open in" button and the image fields all change with it.
  2. Describe the goal specifically. "A cold email offering website redesigns to local gym owners" beats "an email".
  3. Add one line of context. The audience plus one fact the AI can't know is usually enough.
  4. Copy, paste, answer. Use the copy button or open it straight in ChatGPT or Claude, then answer any questions it asks.
  5. Iterate in the chat, not the generator. Once the conversation has context, short follow-ups like "make it shorter" or "more formal" work well.

Want a ready-made starting point instead? Browse the free AI prompt library for tested prompts by category, or read our guide to JSON prompting if you need the AI to return structured data.

Frequently asked questions

Is this AI prompt generator free?

Yes. It is completely free with no sign-up and no limits. It runs in your browser, so there is nothing for us to meter.

Do I need an API key or a ChatGPT account?

You don't need an API key. The generator only writes the prompt; you then paste it into whichever AI you already use, including the free versions of ChatGPT, Claude and Gemini.

Is my text sent anywhere?

No. The prompt is assembled on your device. It only leaves your browser if you click "Open in ChatGPT" or "Open in Claude", which sends it to that service just as if you had pasted it there.

Which AI gives the best results with these prompts?

All the major chatbots handle structured prompts well. Choose the one you already use and let the generator format the prompt for it. If you switch AI later, change the selector and copy the new version.

Why does the Claude version use tags like <task>?

Anthropic recommends XML-style tags for Claude because they mark exactly where each part of the prompt starts and ends. You can type the same tags into your own prompts; they are plain text, not code.

Can I use it for Midjourney?

Yes. Choose Midjourney and the tool switches to image fields and writes a comma-separated prompt with the aspect ratio and exclusions added as parameters at the end, in the format Midjourney's documentation specifies.

Why does the "Open in" button only appear for ChatGPT and Claude?

Those two services accept a prompt in the link, so the button opens a new chat with your prompt already filled in. For Gemini, Midjourney and other tools, copy the prompt and paste it in.

Sources