Generate Your AI Prompts In One Click

What a prompt generator does

A prompt generator takes a short description of a job and expands it into the instructions a model needs to do that job well. You supply the goal. It supplies the structure you would otherwise have to remember every time.

The difference between "write me a cold email" and a prompt that gets a usable cold email is the context the model never sees unless someone writes it down. Who is reading. What they already know. What a good version looks like. What to leave out.

A good generated prompt has five parts, and I check for all five before I trust one:

• A role. Who the model is for this job, stated in one line, with the standard it should work to. "A senior recruiter who has read a thousand cold emails" changes what comes back more than any other single line.
• Context questions. What the model should ask you before it starts: the audience, the product, the tone, the length. A prompt that asks first fails less.
• Steps. The order of work. Draft the subject line last, for example, because it depends on the body.
• An output format. Headings, a table, a numbered list, a character limit. Skip this and you get the model's default, which is usually a wall of bullet points.
• Constraints. What to leave out. Banned words, claims it cannot make, a reading level.

How this generator works

Three steps, and only the first one takes any thought.

1. Describe the goal. One or two lines in plain language. "Summarize customer support tickets into a weekly report for my CEO" is enough. The more specific the line, the less the prompt has to guess.
2. Pick the model. ChatGPT, Claude or Gemini. The generator is tuned for each one, so the prompt is written for the model that will read it.
3. Get the prompt. It appears on the success page as soon as it is ready, and a copy goes to your inbox. Paste it into the model you chose, answer the context questions it asks, and run it.

Each prompt costs $1. If you own the Complete AI Bundle, which is everything I make in one purchase, the generator is free.

Three example inputs and what comes back

I describe the shape here rather than paste full outputs, because the prompt you get depends on the model you pick and the exact words you type.

Input: "Write LinkedIn posts from my podcast transcripts." The prompt sets the role as a ghostwriter for your industry, asks for the transcript, the audience and the posting cadence before it drafts anything, lays out the steps (pull the quotable moments, keep one idea per post, write the hook last), fixes the format at a set number of posts with a character limit each, and lists the words a ghostwriter would ban.

Input: "Turn a messy spreadsheet of leads into a clean CRM import." The prompt puts the model in the role of a data analyst, asks for the column headers and a sample of rows first, then works column by column: normalize names, split addresses, flag duplicates, and output a table in the exact column order your CRM expects. The constraint block tells it to list every row it changed, so you can check the work.

Input: "Help me plan a two-day workshop on AI for my sales team." The prompt sets the role as a workshop facilitator, asks for the team's current tools and the outcome you want on the Monday after, and produces a schedule as a table with time blocks, exercises and the prompt each exercise uses. The format section asks for a materials list at the end, and the constraints keep every session under an hour.

Generator or library: which one to use

The prompt library has 6,900+ ready prompts, sorted by tool and by job, and every one has been through the scoring process described on the documentation page. The generator exists for the job the library does not cover yet.

My rule: search the library first. If the closest prompt needs more than a few edits to fit, describe the job to the generator instead.

How to make a generated prompt better

A generated prompt is a strong first draft. Four edits I make before I rely on one:

1. Add your own examples. Paste two pieces of your real output below the prompt, one you liked and one you did not, and say which is which. The model copies what it sees more faithfully than what it is told.
2. Set the format in your own words. If the prompt says "a report", replace that with the headings you want, in order. Vague format lines produce the model's default structure.
3. Ask it to ask you questions first. Add a line at the end: "Before you start, ask me the five questions whose answers would most change your output." Then answer them. This one edit fixes more bad first drafts than any other.
4. Cut what you will never use. A prompt that carries options you never pick is a prompt the model has to interpret. Delete the branches you do not need.

For the long-form version of this structure, my mega-prompt template is free, and the prompt engineering guide covers the principles behind each of the five parts.