What is Self-critique prompting?
Also called: reflection · critique and revise
Self-critique prompting has the model grade its own draft against named criteria and then produce an improved version — a two-pass loop that beats most single perfect prompts.
The first draft from a model is optimised for plausibility, not for your standard. Asking it to review that draft — "check for unsupported claims, weak structure and anything off-brief, then rewrite" — reliably catches errors and filler the first pass missed. It costs one extra call and is the cheapest quality upgrade available.
The criteria matter: a vague "improve this" produces cosmetic changes; a named rubric produces real fixes.
Now critique your answer against these criteria: accuracy of every claim, completeness versus the brief, clarity for a non-expert. List the weaknesses, then output an improved version.
- Name three to five criteria; make them observable.
- Ask for the weaknesses first, then the rewrite — it prevents rubber-stamping.
- In the Studio, this is the "Check and improve its own draft" toggle.
Use it right now
Ask our brain anything on the homepage — it remembers the whole conversation — or write a brief in the Studio and see the prompt it compiles to.