Show, don't claim
Before / after: what humanizing actually looks like.
Claims are cheap; edits are evidence. Four typical AI drafts, their humanized versions, and exactly what changed — steal the moves for your own text.
Every humanizer site claims results; almost none shows any. Below are honest before/afters: a typical AI draft on the left, a humanized version on the right, and the specific edits explained. The "after" versions were produced with the prompts library and a human pass — the workflow anyone can copy.
Cold email
AI draft
Humanized
What changed and why
- Opener replaced with a specific, checkable reference to the recipient — the single biggest credibility move.
- Every blacklist phrase removed: fast-paced, cutting-edge, leverages, seamlessly, empowering, unlock, delve.
- Vague 'productivity gains' → one concrete number and a named customer.
- The ask got smaller (15 minutes, next week) and an easy no was offered — human senders hedge; bots pitch.
LinkedIn post
AI draft
Humanized
What changed and why
- Listicle-of-nothing replaced with one specific story — specificity is the whole game on LinkedIn.
- All five 'lessons' said nothing; the story teaches one thing memorably.
- Emojis, 'game-changing', 'journey', engagement-bait question: gone.
- A confessed mistake does what 'lead with empathy' claims to do.
Product description
AI draft
Humanized
What changed and why
- Every claim became a number a buyer can verify or picture.
- 'Revolutionary/state-of-the-art/seamlessly/journey' deleted — adjectives replaced by test results.
- The car test detail is the kind of oddly specific fact no template writes — instant authenticity.
- Practical objections (cup holder, cleaning, weight) answered before being asked.
Essay introduction
AI draft
Humanized
What changed and why
- 'Throughout history…' opener — the most flagged phrase in academic AI writing — cut entirely.
- The wheel-to-AI sweep replaced by two specific, citable data points.
- 'In this essay I will explore' → the essay just starts doing it.
- A sharp, arguable thesis replaces 'benefits and drawbacks' mush.
Customer support reply
AI draft
Humanized
What changed and why
- Eight sentences of empathy theater replaced by three facts: what went wrong, what was done, when it lands.
- "Rest assured we are looking into it" → a reference number and a root cause. Reassurance is an artifact; evidence is service.
- A concrete escalation promise with a deadline replaces "your satisfaction is our top priority".
- Note what was kept: apologizing by admitting fault ("that's our error") — one honest sentence outperforms a paragraph of "sincerely apologize for any inconvenience".
Blog introduction
AI draft
Humanized
What changed and why
- "Increasingly popular in recent years" — the classic zero-information opener — replaced by a number and a year.
- A mystery (the two-quarter fall) creates actual reading tension; "in this comprehensive guide we'll explore" creates none.
- The rewrite promises something specific (data, fix, three-line policy) instead of "pros and cons... expert insights".
- First person and an admitted embarrassment do the trust-building that "backed by research" only claims.
The pattern across all six examples
Read the six before/afters in sequence and the same five moves repeat, in every genre. Specifics replace abstractions — a number, a name, a date does the work of three adjectives. The opener states, never warms up — every "before" spends its first sentence clearing its throat; every "after" spends it saying something. One idea beats five — the AI drafts enumerate; the human versions commit. Stakes appear — a mistake admitted, a deadline promised, a mystery unresolved; machine register avoids exposure, human register uses it. Length drops ~30% — not because short is stylish, but because filler was carrying nothing.
That's the whole craft. The blacklist catches the words, the prompts automate the first pass, and this list is the checklist for the final human edit — the five minutes only you can do, because only you know the specifics.
FAQ
Questions.
Were the 'before' examples really AI-generated?
They're representative drafts in the current models' default register — the phrasing patterns (verifiable against the phrase blacklist) are what today's models produce for these tasks by default.
What tool made the 'after' versions?
The prompts from our library plus a human pass for the specific details. The point of the gallery is that the workflow — not a secret tool — does the work.
Why do the after versions have such specific details?
Because specificity is the one thing generic drafts can't fake — a date, a temperature, a named mistake. When you humanize your own text, that detail is your contribution; no rewriter can add facts you didn't give it.
Can I submit my own before/after?
Not yet — a submission option is planned. For now, compare your result against these patterns and the phrase blacklist.