How good is your prompt?
Paste any prompt. Get a score out of 100, feedback that quotes your exact words, the ways it will go wrong, and a rewrite you can copy. Free, 3 grades a day.
What the grader checks
Role and context
Does the prompt tell the model who it is and what it needs to know?
Task clarity
Could a stranger read your prompt and know exactly what to produce?
Constraints and scope
Length, tone, boundaries: the guardrails that stop generic output.
Output format
Structure the result so you can use it without reformatting.
Examples or criteria
Show the model what good looks like, or tell it how it will be judged.
Failure modes
The specific gaps where the model will guess, invent, or drift off task.
Every critique must quote your prompt. If the judge returns feedback without evidence, we reject it instead of showing it to you. That rule is enforced in code, not just in the prompt.
Frequently asked questions
What does the Prompt Grader actually check?
It scores your prompt on five criteria: role and context, task clarity, constraints and scope, output format, and examples or success criteria. Every score comes with a justification that quotes your actual prompt text, a list of the specific ways your prompt is likely to go wrong, and a full rewrite you can copy.
Is it free?
Yes. You get 3 full grades per day free, no signup required. The Pro plan removes the daily cap, keeps your complete grading history, and rewrites your prompt for Claude, ChatGPT, and Gemini in one pass.
Which AI model powers the grading?
Claude Sonnet 4.6 via the Anthropic API, run at temperature 0 for consistent scoring. The grader uses a strict rubric: any critique that fails to quote your prompt verbatim is rejected and re-run rather than shown to you.
How is this different from asking ChatGPT to improve my prompt?
A raw model gives you vague praise and generic advice like "add more detail". The grader enforces a grounding contract: every criterion score must cite a verbatim quote from your prompt as evidence, so the feedback is about your actual words, not boilerplate. You also get a consistent 0 to 100 score you can compare across drafts.
Do you store my prompts?
No. Grading happens in a single API call and your prompt is not saved on our servers. Your grade history lives in your own browser (localStorage) and never leaves your device.
What do the scores mean?
75 or above means the prompt is strong: specific, structured, and ready to use. 45 to 74 means it works but the model is guessing at things you could specify. Below 45 means the model has to invent most of the details, so results will be inconsistent.