vaultaiskillshumanizer
graph

Humanizer

Remove AI tells from prose so it reads like a person wrote it.

created · 2026-05-10
humanizer.md

Humanizer

Strip the visible signature of LLM-generated writing from a piece of text. Use when reviewing your own drafts, editing a teammate's, or rewriting AI output before it goes anywhere a reader will see it.

The point is not to be undetectable. The point is to be honest — text that reads like a real person made it because a real person made the calls.

When to use

  • You wrote a draft alongside an LLM and the seams show.
  • A doc, README, blog post, PR description or product copy needs a final pass.
  • You are reviewing someone else's writing and it feels off but you can't say why.

What to look for

  • Inflated symbolism. "X represents a paradigm shift in how we think about Y." Cut.
  • Promotional voice in neutral contexts. "Robust, scalable, cutting-edge." Replace with concrete claims or delete.
  • Superficial -ing analyses. "Highlighting", "showcasing", "emphasizing" stacked end to end. Pick verbs that do work.
  • Vague attribution. "Some experts argue" — name them or drop the claim.
  • Em dash flood. One per paragraph at most. Otherwise rewrite.
  • Rule of three. Three-item lists where two would do, especially when items are near-synonyms. Compress.
  • AI vocabulary. "Delve", "leverage", "tapestry", "underscores", "in the realm of", "navigate the complexities of". Replace.
  • Negative parallelism. "Not just X, but Y." Most of the time Y alone is the actual sentence.
  • Filler. "It's important to note that…", "In conclusion…", "Ultimately…". Cut.
  • Hedging without commitment. "May potentially be able to help in some cases." Pick a side.

How to apply

  1. Read the whole thing once. Mark anything that triggers the list above.
  2. Rewrite for voice: shorter sentences, concrete subjects, working verbs.
  3. Cut anything that does not advance the argument.
  4. Read it out loud. If your mouth trips, the reader's eye trips.

Output should feel like

  • Specific. Names, numbers, examples.
  • Slightly uneven cadence — humans do not write to a metronome.
  • Comfortable saying "I think" or "I don't know".