Skip to content

humanize-content

Transform AI-generated drafts into authentic-sounding text by removing detectable AI patterns, injecting real voice, and adding personal specificity. The method applies five sequential passes: pattern purge, voice injection, specificity anchoring, structural variation, and read-aloud refinement.

  • Area: Foundational
  • Category: documents
  • Level: beginner
  • Input: conversation
  • Output: document
  • Setup: none
  • Roles: Head of Marketing, Content Manager

Transforms AI-generated text to sound authentically human.

How you ask

  • “Humanize this”
  • “Make this sound less AI”
  • “This sounds robotic”

Inputs

  • text — AI-generated text to humanize (paste or file path)
  • tone-hint — Voice register to write toward when no voice profile is on file. Skill-local — it steers Pass 2 only and is not written to any entity field. (casual | professional | technical)
  • pass-depth — How much of the 5-pass framework to run. Defaults to the full sequence; the reduced setting stops after Pass 1 and is the mandatory setting under 150 words, where Passes 3-5 add more placeholders than prose. Skill-local. (full | pattern-purge-only)
  • text — AI-generated text to humanize
  • writing-samples — Optional 3-5 existing writing samples for voice extraction
  • scope — Whether the profile is the operator's own voice or the company's. The choice decides which profile path the file lands in and which voice later runs load. (personal | company)
  • writing-samples — Optional 3-5 writing samples for voice extraction

Outputs

  • Each rewrite change as one sidecar item (line range, original, suggested, reason, priority, status) the user accepts or rejects per item; schema in content-scorer's references/suggestions-sidecar.md.
  • Each rewrite change as one sidecar item the user accepts or rejects per item, as for quick-humanize.
  • Persistent voice DNA loaded silently by later runs. The scope input picks which of the two paths it lands in.
  • Rewritten text with the change log and the before/after comparison, plus any [YOUR: ...] placeholders the user must fill.
  • Run-scoped voice DNA built from the interview or samples. Temporary — promote it via voice-setup to make it persistent.
  • Voice-anchored rewrite with the per-pass change log and the before/after comparison.
  • prefer a permanent directory)
  • {me.profile-root}/humanize-voice-profile.md