Self-Critique Prompting

Self critique prompting is an advanced technique where the AI is asked to review, evaluate, or improve its own output. Instead of accepting the first answer, the user asks the model to check the response against clear criteria.

This technique is useful for improving quality, clarity, completeness, structure, and tone. It works best when the critique is guided by specific standards rather than a vague request like “make it better.”

What is Self-Critique Prompting?

Self-critique prompting asks the model to act as a reviewer after generating an answer. The model may check whether the output follows instructions, satisfies the audience, includes required details, avoids unsupported claims, or matches the desired format.

Core Idea: Self-critique prompting creates a review step before the final answer is accepted.

When to Use Self-Critique

Writing Improvement
Use it to improve grammar, clarity, flow, tone, and structure.
Prompt Testing
Use it to check whether a prompt produces the desired output.
Report Review
Use it to identify missing sections, weak arguments, or unclear recommendations.
Content Quality
Use it to review whether content fits the audience, objective, and format.

Self-Critique Workflow

Review and Improve Flow

Generate Draft
→
Set Criteria
→
Review Output
→
Identify Gaps
→
Revise

Weak vs Strong Self-Critique Prompts

Weak Prompt Problem Strong Self-Critique Prompt
Review this. Review criteria are missing. Review this blog draft for clarity, beginner-friendliness, structure, examples, and SEO relevance.
Make it better. The model does not know what better means. Identify three weaknesses in this answer and rewrite it to improve accuracy, structure, and practical usefulness.
Check my prompt. The target output is unclear. Check this prompt for missing instruction, context, output format, constraints, and examples.

Useful Critique Criteria

Criterion What It Checks Example Instruction
Accuracy Whether the answer contains unsupported or questionable claims. Flag statements that may need verification.
Completeness Whether all required parts are included. Check whether the answer covers every requested section.
Clarity Whether the answer is easy to understand. Identify sentences that may confuse beginners.
Format Whether the output follows the required structure. Check whether the response follows the requested table format.
Audience Fit Whether the output matches the intended reader. Review whether the language is suitable for first-year students.

Practical Example

Self-Critique Prompt

“Review the previous answer using these criteria: clarity, completeness, beginner-friendliness, examples, and structure. First list the top three issues. Then provide a revised version that fixes those issues.”

This prompt works because it defines how the review should happen and asks for both diagnosis and improvement.

Self-Critique for Prompt Improvement

Self-critique is also useful for improving prompts themselves. You can ask the AI to inspect a prompt and identify missing parts. This helps improve instruction, context, format, examples, and constraints before using the prompt for important work.

Prompt Review Example

“Evaluate this prompt. Tell me what is missing in terms of task clarity, audience, context, output format, constraints, and examples. Then rewrite the prompt in a stronger version.”

Limitations of Self-Critique

Self-critique can improve output quality, but it is not perfect. The model may miss its own mistakes or fail to detect factual errors. For important work, self-critique should be combined with human review and external verification.

Important: Self-critique improves quality control, but it does not replace expert judgment.

High-Risk Mistake: Do not rely on AI self-review alone for legal, medical, financial, academic, or business-critical decisions.

[Image/Diagram: A review loop showing draft, critique criteria, issue detection, revision, and final output.]

Reusable Self-Critique Template

Self-Critique Prompt Template

“Review the output using these criteria: [criteria]. Identify the main weaknesses. Then revise the output to improve [quality goals] while preserving [what should remain unchanged].”

Key Takeaways

  • Self-critique prompting asks AI to review and improve its own output.
  • It works best when clear criteria are provided.
  • Useful criteria include accuracy, clarity, completeness, format, and audience fit.
  • Self-critique can improve drafts, reports, prompts, and content quality.
  • Human review remains necessary for important decisions.