Account

What new tool do you want?

Text & Writing

AI Prompt Generator & Improver

Turn a rough request into a clear, structured prompt for ChatGPT, Claude or Gemini — with role, context, task, constraints and output format.

How to use

  1. 1Describe what you want in plain words, plus who it's for and the format you'd like.
  2. 2Copy the structured prompt into your AI assistant and adjust any details.

How it's calculated

Good prompts state the goal, give context and audience, set constraints (length, tone, what to avoid), specify the output format and invite clarifying questions.

Frequently asked questions

Why do structured prompts work better?

AI models follow explicit instructions. Stating audience, constraints and format removes guesswork and reduces generic answers.

Should I give examples?

Yes — one or two examples of the style or format you want are among the most effective additions to any prompt.

Does this send my text to an AI?

No — the prompt is assembled in your browser from a template. You paste it into the AI tool of your choice.

Worked example: AI Prompt Generator & Improver

The task "write a welcome email for new customers of my online plant shop," with role "an experienced email marketer," audience "first-time plant owners," and 2 versions requested, assembles into a structured prompt starting "You are an experienced email marketer." then "Task: Write a welcome email..." then a Context block listing audience, tone and length as bullet points, then "Output: plain text. Give 2 distinct versions, labeled Version 1–2," and finally a line inviting the AI to ask up to 3 clarifying questions first. Every field you filled in becomes one explicit instruction rather than being buried in a single run-on sentence — which is exactly the difference between a vague request and a prompt that reliably produces what you actually wanted.

Why asking the model to request clarification first is a deliberate default

The closing line — inviting up to 3 clarifying questions before the model proceeds — exists because even a well-structured prompt can omit something that turns out to matter (brand voice specifics, a constraint you forgot to mention), and a model that asks first catches that gap before generating output you'd have to discard and redo. This trades a small amount of back-and-forth for a meaningfully higher chance the first real response is usable, which is a different and often better tradeoff than a prompt engineered to force an immediate answer no matter what's missing.

Why format and constraints do more work than most people expect

Explicitly stating the output format (plain text vs. a Markdown table vs. valid JSON only) changes not just the shape of the response but how literally the model treats everything else — asking for "JSON only, no commentary" specifically suppresses the conversational preamble a model would otherwise add ("Sure, here's..."), which matters a great deal if you're piping the output into another program rather than reading it yourself. Likewise, an explicit length constraint like "under 150 words" and an "avoid: jargon and exclamation marks" line both function as hard guardrails against the single most common failure mode of unstructured prompts: a response that's technically relevant but far too long, too generic, or stylistically wrong for where it's actually going to be used.

Version count is clamped, not just suggested

Entering a number of versions outside the 1–5 range doesn't produce an error — the generator silently clamps it, so typing 12 still only asks the model for 5 distinct versions, and leaving the field blank or non-numeric falls back to 1 rather than breaking the prompt. The upper limit exists because asking a model for too many parallel variations in a single response tends to backfire in practice: each version gets a smaller share of the response to work with, and by the seventh or eighth variant they often collapse into near-duplicates that differ only in word choice, which defeats the point of requesting multiple versions in the first place.

Related tools

More: Text & Writing