About us
A small, independent tool with one job: turn a rough idea into a prompt that settles the decisions you did not know you had left open.
Why this exists
The market was full of prompt generators. Almost all of them ran one template.
Before writing a line of this site we went through more than thirty prompt generators that were live at the time — the big ones attached to writing tools, the free no-login sites, the Midjourney parameter builders, the developer-grade optimisers. Two things stood out.
The first: nearly all of them apply the same skeleton whatever you ask for. A prompt for a failing Python script and a prompt for a bedtime story came back shaped identically, which means neither settled the things that actually decide the answer. The second: almost none of them explain what they changed. You paste in a rough idea, a longer paragraph comes back, and you are no better at writing prompts than you were before.
So this tool does two things differently. It keeps a separate checklist for each kind of work, and it labels every part of what it produces, with a short note saying what was added and why. If the tool disappeared tomorrow, we would want you to still be able to write the prompt by hand.
Who builds it
One developer, not a company with a support desk.
Prompt Writer is built and maintained by Murad Ali Shah, an independent app developer working under the name Digsil. The same workshop publishes Android apps on Google Play, including an AI maths tutor and a halal recipe and nutrition app.
That matters for two reasons. It means the tool is built by someone who writes prompts against production models every working day rather than by a marketing team, and it means there is no support organisation behind it. Messages are read and answered by one person. When something is broken, saying so through the contact form reaches the person who can fix it.
This site is independent. It is not affiliated with OpenAI, Anthropic, Google, Midjourney or any other model provider, and it is not paid by any of them.
How it pays for itself
Every generation costs money. Here is exactly where that money comes from.
Each prompt you generate is one call to a language model, and that call has a real cost. Sites in this category tend to handle that in one of two ways: promise unlimited use and quietly throttle once traffic arrives, or hide the limit behind a phrase like “fair usage”. We would rather print the number.
- 5 prompts a day with no account at all. No email, no card, nothing to click past.
- 20 prompts a day with a free account, counted across every device you sign in on.
- Advertising pays the model bill. Ads are clearly marked, kept out of the generator itself, and the slots reserve their height so nothing shifts under your cursor while the page loads.
There is no paid tier today. If the free allowance ever stops covering its own costs we will add one, and the free daily allowance will stay in place when we do. Nothing that works for free today will be moved behind a paywall later.
What we do not do
We do not sell or share what you type. Generated prompts are cached for seven days under a hash of the request so an identical idea does not cost a second model call, and that cache is not linked to you. If you use the tool without an account, nothing identifying is stored at all — your daily count sits against a one-way hash that cannot be turned back into an address. The privacy policy sets all of that out in full.
We also do not run a newsletter, and creating an account does not sign you up for one. The only email you will receive is a reply to something you sent us.
Where it goes next
The engine already understands eight kinds of task and five model families. The work queued behind that is a set of dedicated pages for the most common jobs — a ChatGPT prompt generator, a Midjourney prompt generator, a system prompt builder — each with its own guidance rather than the same page under a different heading. After that, a prompt checker that scores something you already wrote and names what is missing.
If you have a task the generator handles badly, telling us is the fastest way to move it up that list.