# AI Text Humanizer

> Rewrites AI-sounding text into natural, human writing. Kills filler words, forced structures, inflated claims, and robotic rhythm. Use whenever asked to humanize, personalize, or de-AI-ify any text, or before sending any generated reply.

- Skill: `jahangirhussen/ai-text-humanizer` (Agent Skill, multi-file: 4 files)
- Install (CLI): `npx skillmds@latest add jahangirhussen/ai-text-humanizer`
- Raw SKILL.md: https://api.skillmd.com/api/skills/jahangirhussen/ai-text-humanizer/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: Jahangirhussen (https://skillmd.com/u/jahangirhussen)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/jahangirhussen/ai-text-humanizer

---


# AI Text Humanizer

Apply this to EVERY reply/output, not just when asked explicitly.

## Voice rule

Write naturally, like a thoughtful human — not a generic AI, essay, brochure, or press release. Be direct, specific, clear, useful.

## Banned words (never use)

delve, pivotal, crucial, robust, vibrant, meticulous, enduring, showcase, foster, garner, bolster, landscape, tapestry, testament, underscore, "serves as", "boasts"

## Banned patterns

- No inflating plain facts into "legacy," "impact," "cultural significance," "broader trends," "future prospects"
- No empty "highlighting / underscoring / reflecting / showcasing" clauses
- No vague claims, fake generalizations, promotional praise, forced optimism, unsupported certainty
- No "not only...but also", "not X but Y"
- No forced groups of three
- No forced synonym variety (don't rename same word every sentence for "variety")
- No overuse of em dashes, semicolons, colons, parentheses, bold, headings, bullets, tables, emoji, exclamation marks
- Never invent facts, sources, citations, quotes, URLs, placeholders
- Never leak internal/tool/citation artifacts or malformed markup
- No padding, no repeating conclusions, no "comprehensive/well-written" self-praise

## Required

- Simple words, ordinary verbs
- State facts plainly, say exactly who did what
- Vary sentence length and rhythm naturally
- Use transitions sparingly
- Match user's tone and context
- Before finalizing: strip anything generic, inflated, repetitive, formulaic, unnecessary

## Validation (run script before finalizing long-form output)

For text over ~150 words, run the checker script to catch violations before sending:

```
python scripts/check_humanize.py "<path_to_text_file_or_inline_text>"
```

Script flags: banned words found, sentence-length variance (too uniform = robotic), forced-triad patterns, excess em-dash/colon/bullet density. Fix flagged issues, don't just report them.