Why pick specific goals instead of asking to 'improve' my writing?

'Improve this' is interpreted differently every run and often rewrites your voice away. Naming concrete targets — cut passive voice, reduce jargon, tighten by 20% — produces consistent, reviewable edits that change only what you intended.

What does the preserve list do?

It tells the model which elements are off-limits — direct quotes, defined terms, names, or a distinctive voice. This keeps the edit surgical and prevents the model from flattening intentional style choices.

Should I ask the model to show its changes?

Yes, when you are reviewing rather than batch-processing. A change log or tracked edits lets you accept or reject each revision instead of trusting a wholesale rewrite.

Is my text uploaded anywhere?

No. The builder only assembles a prompt from your selections. Nothing you type is sent to a server.

Writing Improvement Prompt Builder

Writing improvement prompt builder

“Make this better” is the least useful instruction you can give an editor, human or AI — it invites a rewrite that erases your voice. The reliable alternative is to name exactly what you want fixed. This builder turns a set of concrete improvement goals into a precise revision prompt that tells the model which weaknesses to target and which elements to leave untouched.

How it works

You select from a list of specific, well-understood editing targets — improve clarity, cut wordiness, convert passive to active voice, reduce jargon, smooth transitions, fix tone. Each selected goal expands into an explicit instruction in the generated prompt, so the model addresses them one by one rather than guessing. Your preserve list becomes a set of hard constraints — keep direct quotes verbatim, retain defined terms, do not alter the author’s voice — that bound the edit. You can also request the output format, including a change log that explains each revision for review.

Tips and examples

Stack two or three goals, not eight. A focused pass on concision and passive voice beats a scattershot rewrite that touches everything.
Always preserve quotes and terms. If your text cites sources or defines terminology, list them so the model does not paraphrase them away.
Ask for a change log when reviewing. Seeing “cut ‘in order to’ → ‘to’” lets you accept edits with confidence instead of diffing manually.
Run goals in sequence for long pieces. One pass for structure, a second for line-level concision, often produces cleaner results than one mega-prompt.