Examples teach faster than rules. Six common text types, each shown before and after — and one line on what the rewrite actually changed.
Work writing
Before: I hope this message finds you well. I wanted to kindly follow up regarding my previous message about the proposal.
After: Are you still able to review the proposal by Thursday? Happy to walk through it on a call if that is faster.
The edit: request first, an alternative offered, zero ceremony.
Before: This solution enhances productivity by streamlining your daily workflows.
After: The tool cuts weekly reporting from three hours to about forty minutes by drafting the first version from your source sheet.
The edit: a named task and a real number replace two buzzwords.
Before: We are working diligently to ensure continued progress on all deliverables.
After: The draft is 70% complete. Blocker: waiting on Q2 numbers from finance. Review is planned for Friday.
The edit: status, blocker, next step — the three things the reader wanted.
Personal writing
Before: Social media has both positive and negative effects on society that are important to consider.
After: Among teenagers, algorithmic feeds shifted social comparison from classmates to strangers — a change that survey data from 2019–2024 links to declining self-reported wellbeing.
The edit: a narrow claim with scope, mechanism and evidence type.
Before: Excited to share some thoughts on the importance of customer feedback in today's business landscape!
After: A customer asked why exporting a report takes four clicks. We did not have a good answer.
The edit: one real scene replaces an announcement about having thoughts.
Before: Spearheaded cross-functional initiatives to drive operational excellence.
After: Merged two reporting workflows for a 6-person team, cutting monthly close from five days to three.
The edit: action, scope and result — each one checkable in an interview.
The pattern
Look across all six examples: the rewrite is never just smoother. In each case, abstraction got traded for a specific — a task, a number, a scene, a blocker, a mechanism. That specific is something the model cannot know unless you provide it. Which is why the practical workflow is: collect your facts first, then ask for the rewrite with the facts in the prompt. A rewrite prompt without facts can only produce a smoother version of the same vagueness.
Try it: take the weakest paragraph you have, list three facts it should contain but does not, then prompt: “Rewrite this using these facts: [facts]. Keep all existing claims unchanged. Text: [paste]”.
More examples
Less common in example galleries, more common in real work.
Before: We sincerely apologize for any inconvenience you may have experienced. Our team is working diligently to ensure the best possible service.
After: You're right to be frustrated — sync was down for 6 hours on Tuesday. Cause: a failed database migration. Your data is intact. We've added a pre-migration check so this specific failure can't repeat.
The edit: admission, cause, status of their data, prevention. Apology adjectives replaced by information.
Before: When it comes to choosing project management software, there are many factors to consider in today's competitive landscape.
After: For teams under 20 people, the choice usually comes down to two factors: whether you bill by the hour, and whether clients need access.
The edit: the query gets answered in sentence one, with a scoped claim a reader can act on.
Before: We had a productive discussion covering various topics and alignment was achieved on next steps going forward.
After: Decided: ship the pricing change March 1. Open: whether annual plans grandfather in (Lena owns, decides Friday). Parked: the reseller question until Q3.
The edit: decided / open / parked. A recap is a status table wearing prose.
Using examples well
Reading examples feels productive; most of the value evaporates by the next rewrite. What works is extracting the move, not admiring the result. For each pair on this page, name the operation in your own words — “apology adjectives became cause and status”, “landscape opener became a scoped answer” — and you have a reusable tool instead of a memory of someone else's sentence. There are only about a dozen moves behind every example here: delete the ramp, name the actor, scope the claim, attach the number, commit to a verdict, end on consequence. Every good rewrite is two or three of them applied in sequence.
The second use of examples is calibration inside prompts. One before/after pair pasted into a prompt teaches register better than a paragraph of adjectives — “rewrite in the style of this transformation: [before] → [after]” transfers the move to the model directly. Build a private file of your five best pairs from your own work; it becomes both your style guide and your prompt library at once.
And a warning the examples themselves can't show: every rewrite above worked because the underlying facts existed. The support reply needed the cause; the recap needed the decisions. When a rewrite feels impossible, the problem is usually missing information, not missing technique — the honest next step is to go get the fact, not to phrase around its absence.
FAQ
They trade abstraction for specifics: a named task, a real number, a checkable claim. The tone improvement is a side effect of the specificity.
Yes — pasting one before/after pair into your prompt as a style example is one of the most effective ways to steer a rewrite.
Because the decisive input was the facts, not the wording of the prompt. Each rewrite needed a detail the original draft did not contain.
The principles do — cut filler openings, add concrete detail, end with consequence. The specific phrase list differs per language.
Related