Humanizer
Humanizer detects and removes AI writing patterns from any text. It uses 24 pattern detectors, 500+ flagged vocabulary terms, and statistical analysis to identify AI-sounding prose — then rewrites it to sound natural and human.
30,000+ downloads. Use it before publishing anything your agent drafts.
Install:
npx clawhub@latest install humanizerclawhub.ai/brandonwise/ai-humanizer (opens in a new tab)
What it detects
Content patterns: inflated symbolism, superficial analysis, vague attributions, promotional language, false urgency
Language patterns: AI vocabulary words ("delve", "tapestry", "seamless", "leverage", "synergy"), em dash overuse, rule of three, excessive hedging
Style patterns: metronomic sentence length (AI writes in even rhythms; humans don't), low vocabulary diversity, unnaturally high readability scores
Statistical markers: burstiness (humans write in bursts — short sentences, then long ones), type-token ratio, sentence variation
Basic usage
Humanize this article draft: [paste text]Review this blog post for AI patterns. Flag the top 5 most AI-sounding sections and explain why.Rewrite the following paragraph to sound more human. Keep the meaning but vary the sentence length and remove any AI vocabulary: [paste paragraph]Always-on mode
The real power of Humanizer isn't on-demand cleanup — it's baking the principles into how your agent writes by default.
Add this to your ~/SOUL.md (your agent's personality file):
## Writing style
Apply humanizer principles to all writing:
- Vary sentence length — mix short punchy sentences with longer ones
- Avoid: delve, tapestry, seamless, leverage, synergy, robust, holistic, empower, and similar AI vocabulary
- Avoid em dashes as a default connector
- Write with specificity — replace vague attributions ("experts say") with named sources or cut them
- No rule-of-three structure unless it genuinely fits
- Burstiness > consistency — uneven rhythm reads as more humanWith this in SOUL.md, your agent applies humanizer principles to everything it writes — not just when you explicitly invoke the skill.
Run Humanizer before publishing to ZenBin, sending emails via Clawvisor, or posting to Slack. A two-step workflow: draft → humanize → publish. Add this to your morning-brief skill or any content-creation skill.
Score your content
Humanizer can score text for AI detection risk before you publish:
Score this article for AI detection risk on a scale of 1–10. List the specific patterns driving the score.A score of 7+ means it will likely trigger AI detectors. Humanize and rescore until you're below 4.