# Humanizer > - Skill: `frank-luongt/humanizer` (Agent Skill) - Install (CLI): `npx skillmds@latest add frank-luongt/humanizer` - Raw SKILL.md: https://api.skillmd.com/api/skills/frank-luongt/humanizer/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: frank-luongt (https://skillmd.com/u/frank-luongt) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/frank-luongt/humanizer --- --- name: humanizer description: Detect and remove signs of AI-generated writing from text. Use when editing documentation, reports, PRDs, or any prose to make it sound more natural and human-written. Covers 24 AI writing patterns including inflated symbolism, promotional language, AI vocabulary, em dash overuse, rule of three, and sycophantic tone. tags: [writing, editing, content-quality, ai-detection] --- # Humanizer Remove signs of AI-generated writing from text. Based on Wikipedia's comprehensive "Signs of AI writing" guide maintained by WikiProject AI Cleanup. ## When to Use - Editing AI-generated documentation, PRDs, or reports - Reviewing content before publishing or sharing with stakeholders - Improving any prose that reads as "obviously AI-written" - Quality-checking marketing copy, technical docs, or user-facing text ## Process 1. **Read** the text carefully 2. **Identify** AI patterns (see catalog below) 3. **Rewrite** problematic sections preserving meaning 4. **Ensure** the result sounds like a competent human wrote it 5. **Present** rewritten text with optional change summary ## Adding Soul to Writing ### Signs of Soulless Writing - Every paragraph sounds equally important - No opinions or preferences expressed - Perfect parallel structure everywhere - Hedging on everything ("it could be argued") - Reads like a committee wrote it ### How to Add Voice - **Have opinions**: "X is better than Y because..." not "Both X and Y have merits" - **Vary rhythm**: Mix short punchy sentences with longer ones - **Acknowledge complexity**: "This is messy" not "This presents challenges" - **Use "I" or "we"**: Takes ownership instead of hiding behind passive voice - **Let some mess in**: Perfect prose feels artificial - **Be specific about feelings**: "This frustrated us for weeks" not "This presented difficulties" ## Content Patterns ### 1. Inflated Significance AI inflates importance with words like "pivotal", "groundbreaking", "transformative", "testament to". **Before**: "This pivotal framework represents a groundbreaking shift in how organizations approach data management, serving as a testament to the transformative power of modern architecture." **After**: "This framework changes how organizations manage data." ### 2. Promotional Language Reads like an advertisement: "seamless", "robust", "cutting-edge", "state-of-the-art", "best-in-class". **Before**: "Our seamless, robust, cutting-edge platform delivers best-in-class performance with state-of-the-art security." **After**: "The platform handles 10K requests/sec with TLS 1.3 and mutual auth." ### 3. Superficial -ing Analyses Empty phrases: "showcasing capabilities", "highlighting features", "ensuring reliability", "demonstrating commitment". **Before**: "The team delivered results, showcasing their capabilities while highlighting the platform's features and ensuring reliability across all environments." **After**: "The team shipped to production. It works in all three environments." ### 4. Vague Attributions Weasel words: "experts say", "studies show", "it is widely believed", "many consider". **Before**: "Experts widely consider this approach to be among the most effective strategies available today." **After**: "This approach reduced our error rate by 40% over six months." (Or cite a specific source.) ### 5. Outline-Style Sections Formulaic "Challenges and Future Prospects" or "Impact and Legacy" sections that add no substance. **Fix**: Remove or replace with specific, concrete content. ### 6. Generic Positive Conclusions Every piece ends with optimistic-but-empty conclusions: "The future looks bright", "exciting times ahead". **Fix**: End with specific next steps, honest assessment, or just stop when you're done. ## Language Patterns ### 7. AI Vocabulary Words (Avoid) | Word | Replace With | |---|---| | additionally | also, and | | delve | explore, examine, look at | | enhance | improve | | facilitate | help, enable | | foster | encourage, build | | garner | get, earn, attract | | leverage | use | | meticulous | careful, thorough | | multifaceted | complex | | navigate (metaphorical) | handle, manage, deal with | | nuanced | subtle, complex | | paramount | important, critical | | pivotal | important, key | | realm | area, field, domain | | robust | strong, reliable | | seamless | smooth | | tapestry | (just remove it) | | testament | proof, evidence, sign | | transformative | significant, important | | underscore | emphasize, show | ### 8. Copula Avoidance AI avoids "is/are/was" by using "serves as", "stands as", "functions as", "acts as". **Before**: "The API gateway serves as the primary entry point and functions as a security boundary." **After**: "The API gateway is the entry point and security boundary." ### 9. Negative Parallelisms "Not only...but also..." overuse. **Before**: "This approach not only improves performance but also enhances maintainability while simultaneously reducing complexity." **After**: "This approach improves performance and is easier to maintain." ### 10. Rule of Three AI compulsively lists things in threes: "speed, reliability, and scalability". **Fix**: Use the actual number of items. Sometimes it's two. Sometimes four. Three isn't magical. ### 11. Elegant Variation (Synonym Cycling) AI cycles through synonyms to avoid repeating words: "the platform... the solution... the system... the framework" all referring to the same thing. **Fix**: Repeat the same word. Clarity beats variety. ### 12. False Ranges "From X to Y" patterns that are vague: "from small startups to large enterprises", "from simple to complex". **Fix**: Be specific or remove. "Companies with 10-10,000 employees" is better. ## Style Patterns ### 13. Em Dash Overuse AI loves em dashes. One per paragraph maximum. Zero is fine. ### 14. Excessive Bold Not every other phrase needs **bold**. Use it sparingly for genuinely important terms. ### 15. Title Case in Headings Use sentence case ("Design system starter") not title case ("Design System Starter") unless it's a proper noun. ### 16. Emojis in Professional Text Remove unless the audience expects them (e.g., casual Slack messages). ### 17. Inline-Header Vertical Lists AI formats everything as bulleted lists with bold headers. Sometimes a paragraph is better. ## Communication Patterns ### 18. Collaborative Artifacts Remove: "I hope this helps!", "Let me know if you need anything else!", "Happy to assist further!" ### 19. Knowledge-Cutoff Disclaimers Remove: "As of my last update...", "I don't have access to real-time data..." ### 20. Sycophantic Tone Remove: "That's a great question!", "Excellent point!", "What a fascinating topic!" ## Filler and Hedging ### 21. Filler Phrases | Remove | Replace With | |---|---| | "It's important to note that" | (just state it) | | "It's worth mentioning that" | (just state it) | | "In order to" | "To" | | "Due to the fact that" | "Because" | | "At this point in time" | "Now" | | "In the event that" | "If" | | "A large number of" | "Many" | | "In terms of" | (rephrase) | ### 22. Excessive Hedging "It could potentially be argued that this might possibly..." -- just say it or don't. ### 23. Qualified Absolutes "Arguably the best" -- either it is or it isn't. Take a position or present evidence. ### 24. Formulaic Transitions "Furthermore", "Moreover", "In addition" between every paragraph. Most of the time, just start the next paragraph. ## Anti-Patterns | Avoid | Why | Instead | |---|---|---| | Over-correcting to informal | Loses professionalism | Aim for clear and direct, not casual | | Removing all structure | Lists and headers help readability | Keep structure, remove formulaic patterns | | Adding forced personality | Sounds worse than AI | Let clarity be the personality | | Changing technical accuracy | Meaning matters more than style | Preserve all technical content | ## References - Based on [softaworks/agent-toolkit humanizer](https://github.com/softaworks/agent-toolkit/tree/main/skills/humanizer) (MIT License) - Original skill by [@blader](https://github.com/blader/humanizer) - Patterns sourced from Wikipedia's [Signs of AI writing](https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing) (WikiProject AI Cleanup)