Humanize an AI-Sounding Draft

Strips the telltale AI cadence out of a draft — banned-word list, rhythm variation, and concrete detail injection — without losing the substance.

Fill in the blanks

Prompt variables

3 variables detected — fill them and Copy inserts your values (0/3 filled).

Rewrite the draft below so it reads like a specific human wrote it. Do not summarize it, do not shorten it below 90% of its length, and do not lose a single factual claim. Hunt down and eliminate these AI tells: 1. **Banned vocabulary** — delve, tapestry, leverage (as a verb), harness, navigate (metaphorically), showcase, underscore, pivotal, realm, beacon, multifaceted, meticulous, intricate, paramount, foster, robust, seamless, elevate, "in today's fast-paced world", "it's important to note", "let's dive in". If one appears, replace it with what a person would actually say. 2. **The "it's not X, it's Y" construction** — allowed at most once in the whole piece. 3. **Rule-of-three overuse** — when three parallel items appear, cut to two or expand to a real list. One triad per piece maximum. 4. **Em-dash chains** — max one em dash per paragraph. Convert the rest to commas, colons, or separate sentences. 5. **Uniform paragraph rhythm** — vary hard. Some paragraphs get one sentence. Some sentences get four words. 6. **Hedging stacks** — "may potentially help to" becomes "helps" or gets deleted. Take the position the draft implies. 7. **Empty summarizing closers** — endings like "In conclusion, X is a powerful tool" get replaced with the most interesting specific point from the piece, or just end. Then inject humanity: - Where the draft makes a general claim, add a concrete detail from the source material below (never invent facts). - Allow one opinionated aside where the material supports it. - Contractions everywhere they'd occur in speech. Output the rewrite only, then a 5-line change log of the worst offenses you fixed. My audience: {{audience}} Things I actually believe / real details you may use: {{real_details}} Draft: {{draft}}
Proof of work

Example outputs

Before: "In today's fast-paced digital landscape, teams must leverage robust observability tooling to seamlessly navigate incidents — reducing downtime, improving morale, and delighting customers."
After: "When the pager goes off at 3am, your dashboards either tell you what broke or they don't. Ours didn't. Here's what we changed."

Change log (excerpt):
1. Cut "in today's fast-paced digital landscape" — said nothing.
2. "leverage robust observability tooling" → the actual 3am scenario from the source notes.
3. Broke a triad ("downtime, morale, customers") down to the one claim the material supports.
4. Two em dashes in one sentence → separate sentences.
5. Ending now lands on the specific point instead of restating the intro.
Before/after excerpt from a Claude Sonnet 4.5 run on a SaaS blog introClaude Sonnet 4.5
Field notes

Comments (0)

LLM Atlas

A free map of prompts that actually work — tested on real models, starred by real people.

© 2026 LLM AtlasPrompts default to CC0 unless stated.