Get a Critique Before You Ask for an Edit

An editing pattern where you first ask Claude to critique your draft — name its weak spots, unclear passages, and unsupported claims — before you ask it to rewrite anything, so the edit targets real problems instead of rewriting on instinct.

Jump straight to 'improve this' and Claude rewrites everything, including the parts that were fine, in a smoothing pass that can flatten your voice and quietly change your meaning. A critique first separates diagnosis from treatment: you see what's actually wrong, decide which problems are worth fixing, and then direct a focused edit. It keeps you in control of the draft instead of handing it over wholesale.

Drafts you care about and want to keep ownership of — where you'd rather fix specific problems than accept a full rewrite. Especially when the piece is close and you want to know what's holding it back.

Throwaway text where a quick rewrite is fine, or when you explicitly want Claude's version as a fresh take. Critique-first is about protecting a draft you intend to keep as yours.

Critique this draft before editing anything. Tell me: where the argument is weak or unsupported, where it's unclear or drags, where the tone slips, and what a skeptical reader would push back on. Be specific and quote the lines. Don't rewrite yet — I'll decide which issues to fix and how.

You ask for a critique of a landing-page paragraph. Claude flags that the opening buries the benefit and one claim has no proof. You agree with the first, disagree with the second, and ask it to fix only the opening — keeping your voice and your claim, fixing the one real problem.

Ask Claude to critique from a specific reader's perspective — a skeptical investor, a rushed manager, a domain expert — and to name the single objection most likely to make that reader stop. Targeted critique surfaces the one weakness a generic pass would average away.

• Letting the critique slide straight into a rewrite, losing the chance to choose what changes.
• Accepting every critique point as valid — some are matters of taste; you decide.
• Asking for vague feedback ('is this good?') instead of specific, quoted problems you can act on.