AI Prompt Checker
Paste a prompt. Get a 0-100 score, every weak spot named, and three rewrites you can use.
- Home
/ AI Prompt Checker
What the checker looks at
Every prompt is scored on the same seven axes, weighted to add up to 100. Fixed weights mean the same prompt scores the same every time, so you can re-check after an edit and see whether it actually improved.
Goal clarity
20 pointsWhether the outcome you want is actually stated. This is the single biggest source of vague answers.
Audience
15 pointsWho the output is for. Unstated audience is why copy comes back generic.
Format and structure
15 pointsLength, sections, schema. Without it the model picks a shape at random.
Constraints
15 pointsHard rules and exclusions. What the model must not do is as important as what it should.
Context
15 pointsEnough background to do the job. Missing context is why the model guesses and gets it wrong.
Tone and voice
10 pointsExplicit style, or a reference to match. Otherwise you get house-neutral AI voice.
Examples
10 pointsShowing what good looks like. One example moves output quality more than a paragraph of description.
The full breakdown, including how points are awarded inside each axis, is on the prompt quality score page.
A real example
This is an actual report, not a mock-up. Here is what a perfectly ordinary prompt scores, and why.
The prompt
write a follow up email to a client who hasnt responded
What it got right
- Clear intent to write a follow-up email
- Focus on client communication
What it flagged
Critical · Lack of context
Include details about the client's previous interactions and the subject matter of the initial email.
Major · Undefined audience
Specify the type of client (potential, existing) and their industry.
Major · Missing format and tone
Specify whether the email should be formal or informal, and what to include, such as a subject line.
Minor · No example provided
Provide a brief example of a follow-up email that meets the need.
One of the three rewrites
Draft a follow-up email to a potential client who has not responded to my initial proposal regarding our marketing services. The email should be formal, include a subject line, and express my eagerness to discuss their needs. For example, 'Subject: Following Up on Our Marketing Proposal'.
The point is not that the rewrite is longer. It is that every addition maps to a named gap: the audience, the tone, the format, and an example. You can see why it is better, which means you can do it yourself next time.
What your score means
The number is only useful if you know what to do with it. Roughly:
The prompt names a subject but does not say what done looks like. Output will be generic because the model is guessing at the goal, the reader, and the shape all at once.
The most common band by far. You have said what you want but not for whom, in what format, or under what constraints. This is where most of the frustrating output comes from, because the answer is technically responsive and still unusable.
Usually goal, audience, and format are present but constraints or context are thin. Output is close enough to edit rather than rewrite.
The prompt carries enough that the model is not filling gaps on your behalf. Remaining gains come from adding an example rather than more instruction.
A checker is not a prompt generator
These get conflated, and they solve opposite problems. A prompt generator hands you someone else's prompt for a generic task: useful when you have a blank page, useless when you have a specific job with your own context.
A checker starts from the prompt you already wrote. You keep your own context, your own constraints, and your own voice, and it tells you what is missing. If you have ever pasted a prompt from a list of 100 and found it produced something confidently wrong for your situation, that is the difference.
The practical test: if you cannot describe what you want, you want a generator. If you described it and the answer still came back wrong, you want a checker.
Common questions
Is the prompt checker free?
Three full reports a day are free, with no signup and no card. Past that, credit packs start at $2.99 and the credits do not expire. There is no subscription.
What does the score actually mean?
It is the sum of seven weighted axes, out of 100. The same prompt scores the same every time because the rubric is fixed rather than judged by feel. A score in the 40s usually means the goal is clear but the format, constraints, and audience are missing.
Does it work for ChatGPT, Claude, and Gemini?
Yes. The rubric is model-agnostic, because the things that make a prompt weak (no stated goal, no format, no constraints) break every model in the same way.
Can it check image prompts?
Yes. Image prompts are scored on a separate rubric built for composition, lighting, medium, lens, and technical specs rather than the text axes.
What do I get besides the score?
Every weak spot is named and tagged critical, major, or minor with a concrete fix, and you get three rewrites in different styles: concise, detailed, and structured. The score tells you where you are; the rewrites are what you paste.