What this checks
Most disappointing AI replies trace back to a handful of missing pieces in the prompt, not a weak model. This tool reads your prompt and checks it against five things that consistently make replies sharper:
- Role — is there a persona for the model to adopt?
- Task — is there a clear instruction verb (write, summarize, compare…)?
- Output format — did you say what shape you want back (bullets, JSON, a paragraph)?
- Length — is there a target so the reply isn't too short or a wall of text?
- Tone / audience — does the model know who it's writing for?
It also flags a small set of known anti-patterns: contradictory instructions ("brief but detailed"), vague padding ("please be as detailed as possible"), filler phrases that cost tokens for nothing, and heavy pronoun use that leaves the model guessing what "it" refers to.
How the restructuring works
This is not an AI rewrite — nothing here invents new sentences on your behalf. The restructured version below the checklist takes your own words and reorganizes them into labeled sections (Role, Task, Output format, Tone & audience). Where something is missing, you get an editable bracketed placeholder like [Add a role, e.g. "You are an expert technical editor."] instead of a section quietly staying wrong.
Why a "clarity score"
The score is a simple, transparent count: every finding subtracts points, warnings more than tips. It's meant as a quick before/after signal while you edit — watch it climb as you fill in the missing pieces — not a scientific measure of prompt quality.
FAQ
Does this call an AI model? No. Every check is a fixed rule running in your browser. Nothing is uploaded, and no reply is generated — this tool only analyzes your prompt text.
Will the "restructured" version always be usable as-is? Only if every section is filled in. Any bracketed placeholder means you still need to add something before sending it.
Why does it flag "please" or "detailed"? Not because politeness is bad — a single "please" is fine. It flags padding: multiple filler phrases, or vague words like "detailed" and "comprehensive" used without a number or a concrete definition, since those cost tokens without telling the model anything specific.
Is my prompt stored anywhere? No. The tool keeps no history and writes nothing to local storage.