Is Your AI Prompt Getting a D? How to Write Prompts That Score an A

You've been using AI for six months. You feel like you're getting decent results. But here's an uncomfortable question: how would you know if your prompts were bad?

Most people can't answer that. They've never seen a rubric. They compare their outputs to nothing. And so they stay stuck at 'pretty good' when 'great' was two rewrites away.

This article is a grading rubric for prompts. Use it to audit what you're already running — and to build better ones from scratch.

The 5 Dimensions of a Good Prompt

1. Clarity (Does the model know exactly what you want?)

D-grade: "Write me something about onboarding."

A-grade: "Write a 300-word onboarding email for a new enterprise customer who has just signed a 12-month contract. Tone: warm but efficient. Include: what happens in the first 24 hours, who their point of contact is [name], and a single clear next action."

The difference: specificity of format, audience, tone, and constraints. Vague prompts produce average outputs because the model averages across all possible interpretations.

2. Context (Does the model have what it needs to be expert?)

D-grade: "Review this job description."

A-grade: "You are an experienced technical recruiter specializing in early-stage startups. Review this job description for a senior backend engineer role. We are a 15-person seed-stage company competing for talent against Series B companies. Flag anything that will make strong candidates self-select out."

The difference: the A-grade prompt gives the model a persona, a competitive context, and a specific evaluation lens. Context is not fluff — it is the frame that determines what 'good' means.

3. Constraints (Have you told the model what NOT to do?)

D-grade: "Summarize this article."

A-grade: "Summarize this article in exactly 5 bullet points. Each bullet must be one sentence. Do not include any conclusions or recommendations not explicitly stated in the article. Do not use the word 'important'."

The difference: without constraints, the model optimizes for a generic 'good summary.' With constraints, it optimizes for your specific use case.

4. Output Format (Have you specified how the answer should be structured?)

D-grade: "Give me ideas for content."

A-grade: "Generate 10 content ideas for a B2B SaaS blog targeting VP-level buyers. Format as a table with columns: Title, Target persona pain point, Estimated search volume (high/med/low), Content type (listicle/case study/how-to/opinion)."

The difference: specifying format means you get something you can immediately use, not something you have to reformat before it's useful.

5. Iteration Signal (Have you told the model how to improve?)

D-grade: "Make it better."

A-grade: "The first draft is too formal. Rewrite it so it sounds like a smart friend explaining this at a dinner table — no jargon, no hedging, no 'it's important to note that.' Match the energy of the examples below: [paste examples]."

The difference: 'better' means nothing without a direction. Iteration prompts with examples and specific failure modes converge 3–5x faster.

Grade Your Prompts Automatically

Reading a rubric is useful. Getting instant feedback on your actual prompts is better.

PromptLab's free grader scores any prompt on these five dimensions and gives you a letter grade (A–F) with specific, actionable fixes. It takes 10 seconds. No account required.

Paste your prompt into the grader at http://143.198.136.81:8802 — the grade tool is at the /tools/grade endpoint. You might be surprised what you find.

Most people discover they've been writing C-grade prompts and getting C-grade results. The gap between a C and an A is rarely about the AI model. It's almost always the prompt.