Humanize Writing

Humanize Writing is a free agent skill maintained by Scopeful. It teaches an AI coding agent such as Claude Code, Cursor, Windsurf or Codex how to drive this tool correctly, so you do not have to re-explain it every session. Every published Scopeful skill is free and the install command is public, with no sign-in required. Scopeful also tracks hand-verified USD pricing for 39 AI creative tools at https://www.scopeful.org/tools.

Make AI-written text sound human again.

Tags: writing, editing

Reference


name: humanize-writing description: Use this skill whenever the user wants text to sound less AI-generated and more human. Triggers include "humanize this", "this sounds like ChatGPT", "make it sound like me", "remove the AI feel", "rewrite this naturally", "my readers can tell it's AI", or pasting any draft with a request to fix the tone. Strips AI fingerprints (vocabulary, structure, rhythm) while keeping every fact intact.

Humanize Writing

AI text has fingerprints. Readers spot them in two sentences, and so do detectors. This skill removes the fingerprints without touching the facts.

The contract

Pass 1 — Kill the vocabulary tells

Replace on sight (AI-frequency multipliers vs human text are measured, not folklore — "delve" ~48×, "tapestry" ~35×):

Never Use instead
delve into look at, dig into, cover
tapestry, symphony, landscape (metaphorical) cut, or name the actual thing
multifaceted, pivotal, crucial, robust, seamless specific adjective or none
leverage (verb) use
navigate (metaphorical) handle, deal with
unlock, unleash, elevate, empower, supercharge say what actually happens
foster, facilitate, streamline help, run, simplify
in today's digital age / fast-paced world delete the whole clause
it's important to note / worth noting that delete; just say the thing
furthermore, moreover, additionally (sentence openers) And, Also, or nothing
in conclusion / overall / ultimately delete; end on the point itself
game-changer, revolutionize, cutting-edge describe the actual change

Pass 2 — Kill the structural tells

  1. "Not only X but also Y" and "It's not just X, it's Y" — restructure to plain statements. The parallelism reflex is the strongest tell.
  2. Rule-of-three overload ("fast, reliable, and scalable") — humans don't triple every list. Keep one triple per page, max.
  3. Bullet reflex — if the source turned prose into bullets, turn most back into sentences. Bullets only where a reader will scan.
  4. Uniform sentence length — AI writes 15–25 words per sentence, every sentence. Break rhythm on purpose: one 4-word sentence per paragraph. Then a longer one that runs on the way people actually explain things when they're thinking out loud.
  5. Paragraph symmetry — three 3-sentence paragraphs in a row is a tell. Merge one, split another.
  6. Em-dash density — one — maybe — per paragraph. AI sprinkles them everywhere.
  7. The summary ending — cut any final paragraph that restates the piece. End on the last real point.
  8. Hedge stacking — "may potentially help to" → "helps" or "might help". One hedge max.

Pass 3 — Add the human signals

Workflow

  1. Read the draft. List the 5 worst tells found (show the user — this builds trust in the edit).
  2. Rewrite in one pass applying all three passes above.
  3. Diff-check facts: every number, name, date, and claim from the original present and unchanged.
  4. Output the rewrite, then a 3-line summary of what changed. Offer: "Want it closer to your own voice? Paste a sample of your writing."

What this skill refuses

It rewrites style, not authorship. It won't help pass off AI text as human-written where that's deceptive (academic submissions, "written by a human" claims). Say so plainly if asked.