ChatGPT Prompt Engineering: The “AI Critic” Prompt

TL;DR

Use a critic-first sequence: require diagnosis of weaknesses before requesting a rewrite.

Briefing Cornell Notes

Briefing

A three-step “AI critic” prompt can turn ChatGPT into a targeted editor by forcing it to (1) adopt a critic role, (2) identify specific weaknesses in provided text, and (3) rewrite the material using those critiques. The practical payoff is more precise, richer revisions—especially when the input is short, vague, or missing structure—because the model is constrained to diagnose clarity, coherence, organization, grammar, and style before it rewrites.

The prompt’s structure is straightforward. First, it instructs the model to “act as a Critic” and acknowledge the task. Second, it tells the model to “criticize the following” content (inserted into a text field) and “convince me that it can be better,” explicitly prompting step-by-step problem identification. Third, it orders a rewrite: “great points rewrite the text and improve it based on your critique.” In practice, this sequence pushes the model to produce actionable feedback—like what’s unclear, what’s missing, and what would make the writing more compelling—rather than jumping straight to a polished rewrite.

In the first example, a short webpage-style line—“What skills do you need a support engineer”—is treated as the input. The critic flags a “vague title,” noting that it doesn’t clearly indicate what the content will cover. It also points out limited information due to the text being too short, and it calls out the absence of a conclusion when the draft abruptly ends. The rewrite responds directly: it changes the title to “The essential skills for a career in prompt engineering,” adds an introduction, expands key skill areas (including programming, natural language processing, problem-solving, collaboration, and continuous learning), and includes a conclusion.

The second example applies the same workflow to email subject lines aimed at landing an interview at Apple. The critic rejects options for being generic, presumptuous, or cliché—arguing they fail to differentiate the applicant or clearly communicate qualifications. After critique, revised subject lines incorporate stronger value propositions: enthusiasm paired with specificity, role- and skill-focused framing, and more concrete positioning (for example, highlighting a particular expertise in brackets rather than relying on broad “tech enthusiast” language).

A final example uses the prompt to improve a concise story synopsis tied to the Apple TV+ series Severance. The initial synopsis is criticized for lacking character development and detail, following a predictable arc, and offering insufficient conflict and emotional depth. The rewrite addresses those gaps by adding more character and world texture, increasing complexity, and introducing unexpected tension—such as a mysterious colleague appearance that drives the protagonist’s curiosity and investigation. The result is described as more readable and engaging, with a clearer mystery pull.

Overall, the method matters because it converts “rewrite” into a feedback loop: diagnosis first, then revision. That constraint helps produce edits that are not just fluent, but also structurally and strategically improved for clarity, persuasion, and audience fit.

Cornell Notes

The “AI critic” prompt uses a three-step sequence to improve writing quality: the model adopts a critic role, identifies weaknesses in provided text (clarity, coherence, organization, grammar, style, and audience fit), and then rewrites the content using those critiques. In examples, vague or incomplete inputs—like a short skills headline, generic Apple interview subject lines, and a thin Severance synopsis—receive specific feedback before being revised. The approach works best when the input is underdeveloped, because the critic can point out missing structure (like introductions or conclusions) and suggest concrete additions. The rewrite then reflects those issues through clearer titles, expanded details, stronger differentiation, and added conflict or character depth.

How does the three-step “AI critic” prompt change the quality of outputs compared with a straight rewrite request?

It forces a diagnosis-first workflow. Step 1 has the model acknowledge it will act as a critic. Step 2 requires it to “criticize the following” content and “convince” that it can be better, prompting step-by-step identification of problems such as vague titles, missing structure, limited detail, or weak differentiation. Step 3 then orders a rewrite that directly incorporates the critique, so the final text reflects specific fixes rather than generic polishing.

What kinds of problems did the critic identify in the support-engineer skills text, and how did the rewrite respond?

The critic flagged a vague title (“What skills do you need a support engineer”) that didn’t clearly indicate what the content would cover. It also noted the text was too short to provide enough information, and it pointed out the absence of a conclusion because the draft abruptly ended. The rewrite addressed these by changing the title to “The essential skills for a career in prompt engineering,” adding an introduction, expanding the skill list (programming, natural language processing, problem-solving, collaboration, continuous learning), and including a conclusion.

Why were the initial Apple email subject lines criticized, and what improvements appeared after rewriting?

The critic rejected subject lines for being generic, presumptuous, or cliché—failing to differentiate the applicant or clearly communicate unique qualifications. For example, one line was called generic because it didn’t highlight standout experience or skills, and another was criticized as overly direct and pushy. The rewritten versions improved differentiation by incorporating stronger value propositions, including more specific role/skill framing (such as using brackets to specify the exact position or expertise) and emphasizing relevant achievements or collaboration fit.

What weaknesses did the critic find in the Severance synopsis, and what changes made the revised version more engaging?

The critic said the synopsis lacked character development and detail, followed a predictable plot arc, and had insufficient conflict and emotional depth. It also suggested adding world-building and unexpected plot twists. The revised synopsis responded by adding more character and complexity, increasing conflict, and introducing a mystery driver—like a mysterious colleague appearing outside the work—which pushes the protagonist to uncover the truth.

In these examples, what does “specific changes” usually mean in practice?

It typically means concrete edits tied to the critique: rewriting titles to be more specific, adding missing sections (introductions or conclusions), expanding underdeveloped content with concrete details (skills, examples, or learning paths), and adjusting language to avoid generic or cliché phrasing. In the email case, it also means tailoring subject lines to communicate a clearer value proposition and role fit rather than relying on broad enthusiasm.

Review Questions

Pick a short paragraph you’ve written. Use the critic-first workflow: what clarity, structure, or differentiation issues would a critic likely flag before any rewrite?
Compare two subject lines: one generic and one role-specific. What exact critique points would justify rewriting the generic one?
Take a synopsis or outline with a predictable arc. What types of conflict, character detail, or twist would you add to address the likely critique categories?

Key Points

1
Use a critic-first sequence: require diagnosis of weaknesses before requesting a rewrite.
2
Insert the target text into a dedicated “criticize the following” step to keep feedback grounded in the actual content.
3
Expect the critic to focus on clarity, coherence, organization, grammar, and style, plus audience and purpose fit.
4
Short or incomplete drafts benefit most because the critic can identify missing elements like introductions or conclusions.
5
For outreach subject lines, prioritize differentiation: avoid generic enthusiasm and presumptuous phrasing.
6
For story synopses, address engagement levers—character development, conflict, unpredictability, and emotional depth—before rewriting.
7
Iterate: run the critique and rewrite loop repeatedly until the revised output matches the desired tone and specificity.

Highlights

The prompt’s power comes from forcing a feedback loop: critique first, rewrite second.

In the support-skills example, the critic pinpointed a vague title, missing conclusion, and insufficient detail—then the rewrite added structure and expanded skill categories.

For Apple interview subject lines, the critic rejected generic/cliché options and pushed toward specificity, including role- and skill-focused framing.

In the Severance synopsis, the critic targeted predictability and low conflict, and the rewrite increased mystery and character-driven momentum.

Topics

Prompt Engineering
AI Critique
Text Rewriting
Email Subject Lines
Story Synopsis