The Critique and Refine Pattern
Jay Banlasan
The AI Systems Guy
tl;dr
Ask AI to critique its own output, then refine it. Two-pass prompting produces higher quality results.
The critique refine pattern ai users should adopt is simple: after AI generates something, ask it to critique its own work, then fix the problems it found. Two passes produce significantly better output than one.
This works because AI is often better at evaluating quality than producing it on the first try.
How It Works
Step 1: Generate the output. Ask AI to write the email, the report, the ad copy, whatever you need.
Step 2: Ask AI to critique its own output. "Look at what you just wrote. Identify three weaknesses, any inaccurate claims, and anything that could be clearer."
Step 3: Ask AI to fix the issues it identified. "Now rewrite the piece, addressing all three weaknesses you identified."
The second version is almost always better than the first. Not because AI was holding back. Because the critique step forces it to evaluate from a different perspective.
Why This Beats Single-Pass Prompting
When AI generates text, it optimizes for completion. When it critiques, it optimizes for quality. These are different modes of thinking.
The generation mode fills the page. The critique mode challenges what is on the page. Using both gives you the benefit of AI's breadth (generation) and depth (evaluation).
Advanced Applications
Multi-perspective critique: "Critique this sales email from the perspective of a busy CEO who receives 50 emails a day. Then critique it from the perspective of a sales coach. Then rewrite it addressing both critiques."
Scoring critique: "Rate each paragraph of this landing page copy on a scale of 1-10 for clarity, persuasiveness, and specificity. Rewrite any paragraph below 7."
Competitive critique: "Here is a competitor's approach to this same topic. How does our version compare? What do they do better? Rewrite ours to close any gaps."
The Practical Template
For any important output, use this pattern:
[Initial generation prompt]
Now critique this output:
1. What are the three biggest weaknesses?
2. Is anything inaccurate or misleading?
3. What would make a skeptical reader stop reading?
Now rewrite it, addressing every issue you identified.
When to Use It
Use the critique refine pattern ai produces for anything that matters: client-facing documents, ads, emails to prospects, proposals, and content. Skip it for internal notes and quick tasks where the first pass is good enough.
The extra 60 seconds of prompting saves 10 minutes of editing. And often catches problems you would have missed.
Build These Systems
Ready to implement? These step-by-step tutorials show you exactly how:
- How to Build AI Quality Scoring Pipelines - Automatically score AI output quality to route low-quality results for re-processing.
- How to Implement Output Validation for AI Responses - Build validation layers that catch hallucinations and format errors automatically.
- How to Build AI Evaluation Pipelines - Automate quality scoring of AI outputs using rubrics and judge models.
Want this built for your business?
Get a free assessment of where AI operations can replace overhead in your company.
Get Your Free Assessment