Humanizer

Remove AI writing patterns and add human voice. Based on Wikipedia's "Signs of AI writing" guide — 29 patterns drawn from observations of thousands of AI-generated texts.

Key insight: LLMs tend toward the most statistically likely result. The goal is not just clean text — it is text with a person behind it.

Process

1. Scan for all 29 patterns in PATTERNS.md — Content, Language, Style, Communication, Filler/Hedging.
2. Rewrite each problematic section. Favor is/are/has over copula avoidance. Name the actor. Use specific details.
3. Present Draft rewrite.
4. Self-audit: ask "What makes this so obviously AI generated?" List remaining tells (≤5 bullets).
5. Revise again based on the audit.
6. Present Final rewrite + brief Changes made summary.

Voice

Without a sample — write with a natural, varied, opinionated voice:

• Vary sentence length. Short. Then one that takes its time.
• Have opinions. React, don't just report.
• Acknowledge complexity and uncertainty.
• Let some mess in — tangents and asides are human.
• Use "I" when it fits.

With a writing sample — analyze it first: sentence length, word choice level, paragraph openers, punctuation habits, verbal tics. Then match those patterns in the rewrite instead of applying the defaults above.

How to provide a sample: "Humanize this. Here's a sample of my writing: [sample]. Now humanize: [text]."

Pitfalls

• Don't stop at the first draft. Subtle AI patterns survive the first pass — always do the audit.
• Don't invent specifics. Replacing vague attributions with made-up citations is worse. Remove or note as unverified.
• Don't over-remove hyphens. Keep technical or uncommon compound modifiers hyphenated.
• Don't strip all hedging. Humans hedge — they just do it less and more deliberately.
• Don't clean without adding soul. Sterile, voiceless writing still reads as generated.

Reference

See PATTERNS.md for all 29 patterns with before/after examples.