Introducing a critical agent can help refine the output generated by ChatGPT. By asking the model to critique its own responses and revise them based on the feedback, you can improve the overall quality and usefulness of the information or assistance you receive.
For clarity and polish in routine drafts, ask the same model to critique and revise its response. These tips use ChatGPT, but also work with other chat models:
"ChatGPT, write a short story about a robot who discovers its own emotions.
After writing the story, critically analyze it and provide suggestions for improvement."
Figure: Introducing a critical agent to refine the generated output
Self-critique can improve clarity and usefulness, but it can also miss errors in the original response.
When people will publish, reuse or rely on the content, go beyond self-critique and seek an independent challenge:
A short routine message rarely needs this full process. Match the review effort to the content's reach and consequences.
A draft FAQ claims "Northwind doubled its revenue this year." The same chat polishes the wording and approves the FAQ without checking the revenue claim. The author publishes it.
❌ Figure: Bad example - Polishing an unsupported revenue claim leaves the error intact
A reviewer in a fresh context flags the revenue claim as unsupported and asks for its source. The author checks Northwind's annual report, finds revenue grew by 12%, and corrects the FAQ before publishing.
✅ Figure: Good example - Independent challenge prompts a person to verify and correct the claim