10 Real Prompts I Rewrote — Before and After (With Results)

Most prompt guides show you the finished product. Here's the messy original and why the rewrite works.

Every prompt engineering guide shows you polished, perfect prompts. None of them show you the original garbage prompt that someone actually typed first. That's the part that teaches you something — seeing the transformation from "vague thing I'd naturally type" to "structured prompt that gets great results." Here are 10 real before-and-after examples with explanations.

Key Takeaway

The gap between a bad prompt and a good one is usually 30 seconds of additional context. The AI doesn't need more words — it needs the RIGHT words: role, format, constraints.

1. The Email

Before: "Write me an email to my boss about being late"

After: "Act as a professional communicator. Draft a 3-4 sentence email to my manager explaining that I'll be 30 minutes late due to a traffic delay. Tone: respectful and brief, not overly apologetic. Include an ETA and offer to make up the time. Subject line included."

What changed: Added role, specific length, reason, tone guidance, and format requirements. The "before" prompt gives ChatGPT 50 possible emails to write. The "after" prompt narrows it to one good one.

2. The Code Review

Before: "Review my code"

After: "Act as a senior Python developer conducting a code review. Review the function below for: bugs, performance issues, readability, and Python best practices. For each issue found, explain why it's a problem and provide a corrected version. Format as a numbered list. [paste code]"

What changed: Specified the review criteria (not just "is it good?"), requested actionable fixes (not just identification), and set the output format.

3. The Data Analysis

Before: "Analyze this data"

After: "I'm uploading a CSV of monthly sales data for 2024-2025. Analyze it and provide: (1) Overall trend with percentage change year-over-year, (2) Top 3 performing months and why they might have been strong, (3) Any anomalies or concerning patterns, (4) A 3-sentence executive summary I can paste into a status report. Use specific numbers from the data."

What changed: Told the AI what "analyze" means to YOU. Without this, "analyze" could mean anything — statistical tests, a chart, a summary, a recommendation.

4. The Meeting Summary

Before: "Summarize this meeting"

After: "Summarize the following meeting transcript into: (1) Key decisions made (bulleted list), (2) Action items with owner and deadline, (3) Open questions that still need answers. Keep it under 200 words. If a decision or owner is unclear from the transcript, flag it explicitly."

What changed: Defined what "summary" means in a business context. The AI's default summary is a narrative paragraph. You want structured, actionable output.

5. The Job Description

Before: "Write a job description for a data analyst"

After: "Write a job description for a mid-level data analyst at a 200-person SaaS company. The role focuses on product analytics — analyzing user behavior, running A/B tests, and building dashboards in Looker. Required skills: SQL, Python, and a BI tool. Nice to have: experiment design. Tone: direct and specific, not corporate boilerplate. Avoid phrases like 'rockstar' or 'ninja.' Length: 400-500 words."

What changed: Context (company size, industry), specificity (product analytics, not general analytics), tool requirements, tone constraints, and anti-patterns to avoid.

Pro tip

The pattern in every transformation: the "after" prompt answers three questions the "before" prompt doesn't — WHO is this for, WHAT format do you want, and WHAT should it NOT include.

6. The Social Post

Before: "Write a LinkedIn post about AI"

After: "Write a LinkedIn post (150-200 words) about how I used AI to cut my weekly report time from 3 hours to 20 minutes. Tone: conversational, first-person, genuine — not salesy or preachy. Start with a hook that stops scrolling. End with a question to drive comments. No hashtags in the body text, add 3-5 at the end."

7. The Customer Response

Before: "Reply to this angry customer"

After: "Draft a customer support reply to the complaint below. Acknowledge their frustration without being condescending. Explain what happened (our system had a billing error). Offer a concrete resolution (full refund + 1 month free). Keep it under 100 words. Professional but warm. [paste complaint]"

8. The Excel Formula

Before: "Help me with an Excel formula"

After: "I have an Excel spreadsheet where column A has employee names, column B has departments, and column C has salaries. Write a formula for cell D1 that calculates the average salary for the 'Engineering' department only. Use AVERAGEIF. Explain what each part of the formula does."

9. The Presentation Outline

Before: "Make me a presentation about Q3 results"

After: "Create a 10-slide presentation outline for Q3 2025 business results. Audience: C-suite executives (keep it strategic, not tactical). Slide 1: title. Slide 2: executive summary (3 bullet max). Slides 3-7: key metrics with context. Slide 8: challenges. Slide 9: Q4 priorities. Slide 10: discussion questions. For each slide, write the headline and 2-3 bullets."

10. The Research

Before: "Tell me about competitor analysis"

After: "Act as a business strategist. Create a competitive analysis framework I can use to compare 3 SaaS tools in the project management space (Asana, Monday, ClickUp). Include: pricing comparison, key feature differences, target audience, market positioning, and strengths/weaknesses. Format as a comparison table followed by a 200-word strategic summary recommending which to position against."