Self-Critique Prompts: AI as Its Own Reviewer

Asking the model to critique and revise its own output is one of the cheapest quality boosts in prompt engineering. Master the patterns and their limits.

34 min · Reviewed 2026

Why self-critique works

Generating text and evaluating text are slightly different cognitive operations. When a model is asked to critique an answer (especially without being told it's its own answer), it can spot flaws it missed when generating. This is the foundation behind Constitutional AI and many production pipelines.

Single-prompt self-critique

Task: Write a short explanation of why inflation happens, aimed at a 9th-grade audience.

Step 1: Draft your answer in <draft> tags.
Step 2: In <critique> tags, evaluate the draft:
  - Is it accurate? Flag any oversimplifications.
  - Is it level-appropriate? Point out any jargon.
  - Is it complete? Name one thing missing.
Step 3: In <final> tags, rewrite the draft fixing the issues raised.

Only the <final> block will be shown to the student.A single-prompt self-critique — all three stages in one call.

This works because Claude will genuinely evaluate its own draft when asked. The <final> block is almost always better than the <draft>, often markedly so. The cost is extra tokens (1.5-2x) for significantly higher quality.

Separate-turn self-critique

TURN 1:
Write a resignation letter. (no mention of critique)

TURN 2:
You are a no-nonsense HR consultant. The letter below was written by an employee. Identify the five biggest problems and rewrite it.

<letter>
{RESPONSE_FROM_TURN_1}
</letter>The AI has no idea it's critiquing itself. This sometimes produces harsher, better critique than self-aware critique.

Self-consistency via sampling

A related technique: generate N answers at temperature > 0, then use the model (or majority vote) to pick the best. On reasoning tasks, 5 samples with majority vote on the final answer dramatically outperforms a single sample.

# Pseudocode
answers = []
for i in range(5):
    answer = model.complete(prompt, temperature=0.7)
    answers.append(extract_final_answer(answer))

best = majority_vote(answers)  # or ask a judge modelSelf-consistency with majority voting.

Limits of self-critique

If the model is wrong in a systematic way, its critique will be wrong too.
Self-critique can INCREASE verbosity without improving substance — watch for that.
On factual accuracy, self-critique helps less than retrieval. The model can't fact-check what it doesn't know.
Sycophancy: models sometimes praise their own drafts. Neutral critique framings help.

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-prompting-self-critique-creators

What is the core idea behind "Self-Critique Prompts: AI as Its Own Reviewer"?
1. Asking the model to critique and revise its own output is one of the cheapest quality boosts in prompt engineering. Master the patterns and their limits.
2. Show before/after diffs when requested
3. 'Give a paragraph, no longer.'
4. If you're parsing model output in code, format reliability matters as much as co…
Which term best describes a foundational idea in "Self-Critique Prompts: AI as Its Own Reviewer"?
1. self-consistency
2. self-critique
3. reflection prompting
4. Show before/after diffs when requested
A learner studying Self-Critique Prompts: AI as Its Own Reviewer would need to understand which concept?
1. self-critique
2. reflection prompting
3. self-consistency
4. Show before/after diffs when requested
Which of these is directly relevant to Self-Critique Prompts: AI as Its Own Reviewer?
1. self-critique
2. self-consistency
3. Show before/after diffs when requested
4. reflection prompting
Which of the following is a key point about Self-Critique Prompts: AI as Its Own Reviewer?
1. If the model is wrong in a systematic way, its critique will be wrong too.
2. Self-critique can INCREASE verbosity without improving substance — watch for that.
3. On factual accuracy, self-critique helps less than retrieval.
4. Sycophancy: models sometimes praise their own drafts. Neutral critique framings help.
Which of these does NOT belong in a discussion of Self-Critique Prompts: AI as Its Own Reviewer?
1. Self-critique can INCREASE verbosity without improving substance — watch for that.
2. If the model is wrong in a systematic way, its critique will be wrong too.
3. On factual accuracy, self-critique helps less than retrieval.
4. Show before/after diffs when requested
What is the key insight about "Don't stack critique endlessly" in the context of Self-Critique Prompts: AI as Its Own Reviewer?
1. Show before/after diffs when requested
2. 'Give a paragraph, no longer.'
3. Two critique passes usually beats one. Three+ often produces diminishing returns and can over-polish text into blandness.
4. If you're parsing model output in code, format reliability matters as much as co…
What is the key insight about "Related reading" in the context of Self-Critique Prompts: AI as Its Own Reviewer?
1. Show before/after diffs when requested
2. 'Give a paragraph, no longer.'
3. If you're parsing model output in code, format reliability matters as much as co…
4. Look up 'Reflexion' (Shinn et al. 2023), 'Self-Refine' (Madaan et al. 2023), and Anthropic's Constitutional AI paper.
What is the recommended tip about "Practitioner tip" in the context of Self-Critique Prompts: AI as Its Own Reviewer?
1. Treat every prompt as a spec: role → context → task → format. Review your first output as a draft, not a final.
2. Show before/after diffs when requested
3. 'Give a paragraph, no longer.'
4. If you're parsing model output in code, format reliability matters as much as co…
Which statement accurately describes an aspect of Self-Critique Prompts: AI as Its Own Reviewer?
1. Show before/after diffs when requested
2. Generating text and evaluating text are slightly different cognitive operations.
3. 'Give a paragraph, no longer.'
4. If you're parsing model output in code, format reliability matters as much as co…
What does working with Self-Critique Prompts: AI as Its Own Reviewer typically involve?
1. Show before/after diffs when requested
2. 'Give a paragraph, no longer.'
3. This works because Claude will genuinely evaluate its own draft when asked.
4. If you're parsing model output in code, format reliability matters as much as co…
Which of the following is true about Self-Critique Prompts: AI as Its Own Reviewer?
1. Show before/after diffs when requested
2. 'Give a paragraph, no longer.'
3. If you're parsing model output in code, format reliability matters as much as co…
4. A related technique: generate N answers at temperature > 0, then use the model (or majority vote) to pick the best.
Which best describes the scope of "Self-Critique Prompts: AI as Its Own Reviewer"?
1. It focuses on Asking the model to critique and revise its own output is one of the cheapest quality boosts in prom
2. It is unrelated to prompting workflows
3. It applies only to the opposite beginner tier
4. It was deprecated in 2024 and no longer relevant
Which section heading best belongs in a lesson about Self-Critique Prompts: AI as Its Own Reviewer?
1. Show before/after diffs when requested
2. Single-prompt self-critique
3. 'Give a paragraph, no longer.'
4. If you're parsing model output in code, format reliability matters as much as co…
Which section heading best belongs in a lesson about Self-Critique Prompts: AI as Its Own Reviewer?
1. Show before/after diffs when requested
2. 'Give a paragraph, no longer.'
3. Separate-turn self-critique
4. If you're parsing model output in code, format reliability matters as much as co…

← Back to interactive lesson

Tendril · Creators · Prompting

Self-Critique Prompts: AI as Its Own Reviewer

Asking the model to critique and revise its own output is one of the cheapest quality boosts in prompt engineering. Master the patterns and their limits.

34 min · Reviewed 2026

Why self-critique works

Single-prompt self-critique

Task: Write a short explanation of why inflation happens, aimed at a 9th-grade audience.

Step 1: Draft your answer in <draft> tags.
Step 2: In <critique> tags, evaluate the draft:
  - Is it accurate? Flag any oversimplifications.
  - Is it level-appropriate? Point out any jargon.
  - Is it complete? Name one thing missing.
Step 3: In <final> tags, rewrite the draft fixing the issues raised.

Only the <final> block will be shown to the student.A single-prompt self-critique — all three stages in one call.

Separate-turn self-critique

TURN 1:
Write a resignation letter. (no mention of critique)

TURN 2:
You are a no-nonsense HR consultant. The letter below was written by an employee. Identify the five biggest problems and rewrite it.

<letter>
{RESPONSE_FROM_TURN_1}
</letter>The AI has no idea it's critiquing itself. This sometimes produces harsher, better critique than self-aware critique.

Self-consistency via sampling

# Pseudocode
answers = []
for i in range(5):
    answer = model.complete(prompt, temperature=0.7)
    answers.append(extract_final_answer(answer))

best = majority_vote(answers)  # or ask a judge modelSelf-consistency with majority voting.

Limits of self-critique

If the model is wrong in a systematic way, its critique will be wrong too.
Self-critique can INCREASE verbosity without improving substance — watch for that.
On factual accuracy, self-critique helps less than retrieval. The model can't fact-check what it doesn't know.
Sycophancy: models sometimes praise their own drafts. Neutral critique framings help.

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-prompting-self-critique-creators

What is the core idea behind "Self-Critique Prompts: AI as Its Own Reviewer"?
1. Asking the model to critique and revise its own output is one of the cheapest quality boosts in prompt engineering. Master the patterns and their limits.
2. Show before/after diffs when requested
3. 'Give a paragraph, no longer.'
4. If you're parsing model output in code, format reliability matters as much as co…
Which term best describes a foundational idea in "Self-Critique Prompts: AI as Its Own Reviewer"?
1. self-consistency
2. self-critique
3. reflection prompting
4. Show before/after diffs when requested
A learner studying Self-Critique Prompts: AI as Its Own Reviewer would need to understand which concept?
1. self-critique
2. reflection prompting
3. self-consistency
4. Show before/after diffs when requested
Which of these is directly relevant to Self-Critique Prompts: AI as Its Own Reviewer?
1. self-critique
2. self-consistency
3. Show before/after diffs when requested
4. reflection prompting
Which of the following is a key point about Self-Critique Prompts: AI as Its Own Reviewer?
1. If the model is wrong in a systematic way, its critique will be wrong too.
2. Self-critique can INCREASE verbosity without improving substance — watch for that.
3. On factual accuracy, self-critique helps less than retrieval.
4. Sycophancy: models sometimes praise their own drafts. Neutral critique framings help.
Which of these does NOT belong in a discussion of Self-Critique Prompts: AI as Its Own Reviewer?
1. Self-critique can INCREASE verbosity without improving substance — watch for that.
2. If the model is wrong in a systematic way, its critique will be wrong too.
3. On factual accuracy, self-critique helps less than retrieval.
4. Show before/after diffs when requested
What is the key insight about "Don't stack critique endlessly" in the context of Self-Critique Prompts: AI as Its Own Reviewer?
1. Show before/after diffs when requested
2. 'Give a paragraph, no longer.'
3. Two critique passes usually beats one. Three+ often produces diminishing returns and can over-polish text into blandness.
4. If you're parsing model output in code, format reliability matters as much as co…
What is the key insight about "Related reading" in the context of Self-Critique Prompts: AI as Its Own Reviewer?
1. Show before/after diffs when requested
2. 'Give a paragraph, no longer.'
3. If you're parsing model output in code, format reliability matters as much as co…
4. Look up 'Reflexion' (Shinn et al. 2023), 'Self-Refine' (Madaan et al. 2023), and Anthropic's Constitutional AI paper.
What is the recommended tip about "Practitioner tip" in the context of Self-Critique Prompts: AI as Its Own Reviewer?
1. Treat every prompt as a spec: role → context → task → format. Review your first output as a draft, not a final.
2. Show before/after diffs when requested
3. 'Give a paragraph, no longer.'
4. If you're parsing model output in code, format reliability matters as much as co…
Which statement accurately describes an aspect of Self-Critique Prompts: AI as Its Own Reviewer?
1. Show before/after diffs when requested
2. Generating text and evaluating text are slightly different cognitive operations.
3. 'Give a paragraph, no longer.'
4. If you're parsing model output in code, format reliability matters as much as co…
What does working with Self-Critique Prompts: AI as Its Own Reviewer typically involve?
1. Show before/after diffs when requested
2. 'Give a paragraph, no longer.'
3. This works because Claude will genuinely evaluate its own draft when asked.
4. If you're parsing model output in code, format reliability matters as much as co…
Which of the following is true about Self-Critique Prompts: AI as Its Own Reviewer?
1. Show before/after diffs when requested
2. 'Give a paragraph, no longer.'
3. If you're parsing model output in code, format reliability matters as much as co…
4. A related technique: generate N answers at temperature > 0, then use the model (or majority vote) to pick the best.
Which best describes the scope of "Self-Critique Prompts: AI as Its Own Reviewer"?
1. It focuses on Asking the model to critique and revise its own output is one of the cheapest quality boosts in prom
2. It is unrelated to prompting workflows
3. It applies only to the opposite beginner tier
4. It was deprecated in 2024 and no longer relevant
Which section heading best belongs in a lesson about Self-Critique Prompts: AI as Its Own Reviewer?
1. Show before/after diffs when requested
2. Single-prompt self-critique
3. 'Give a paragraph, no longer.'
4. If you're parsing model output in code, format reliability matters as much as co…
Which section heading best belongs in a lesson about Self-Critique Prompts: AI as Its Own Reviewer?
1. Show before/after diffs when requested
2. 'Give a paragraph, no longer.'
3. Separate-turn self-critique
4. If you're parsing model output in code, format reliability matters as much as co…

← Back to interactive lesson