Strips AI tells from any text
before it ships to humans
A professional editor skill that rewrites filler openers, vague power words, and corporate padding out of AI-generated copy — leaving clean, natural-sounding text with every factual claim intact.
Recommended Models
This skill runs well on any capable model. Pick based on your volume and budget.
| Model | Best For | Cost | Quality | |
|---|---|---|---|---|
| Claude Sonnet 4.6 | Best quality — catches subtle structural tells | Moderate | ★★★★★ | Recommended |
| GPT-5.4 Mini | Great budget option for bulk editing passes | Cheap | ★★★★★ | Recommended |
| Gemini 2.5 Flash | Fastest — ideal for high-volume automation | Cheap | ★★★★★ | Recommended |
| Llama 4 Maverick | Free via Groq — good for personal use | Free | ★★★★★ |
Temperature tip: Set to 0.4. Low enough to follow the editing rules reliably, high enough to produce varied, natural-sounding rewrites instead of robotic substitutions.
Cost Estimate
💚 Cheap Tier
~$0.002
per output, on GPT-5.4 Mini or Gemini 2.5 Flash
~800
input tokens / run
~800
output tokens / run
$2.00
per 1,000 outputs
Runs on any model. Gemini 2.5 Flash keeps this near-free at scale — roughly $0.25 per 1,000 passes ($0.25/M). Use Claude Sonnet 4.6 for higher quality at ~$1.50 per 100 passes ($15/M out).
Before & After Examples
See what the humanizer does to real AI-generated text in the wild.
Email opener
Before — AI output
I hope this email finds you well! I wanted to reach out to discuss a potential opportunity that I believe could be a game-changer for your business. Our cutting-edge solution seamlessly integrates with your existing workflow to unlock significant value.
After — Humanized
Quick one — we built something that might save your ops team about four hours a week. It plugs into the tools you're already using. Worth a 20-minute call?
Before — AI output
In today's rapidly evolving digital landscape, businesses must leverage cutting-edge AI tools to remain competitive. It's worth noting that the biggest takeaway from recent research is that organisations that fail to adapt risk being left behind.
After — Humanized
Most companies using AI are still running it like a fancy search engine. The ones pulling ahead are doing something different — they're building reusable skills that compound. Here's what that looks like.
Model Compatibility
✓ Works Best With
- Claude Sonnet — highest quality rewrites, catches structural tells
- GPT-5.4 Mini — great bulk-editing throughput at low cost
- Gemini 2.5 Flash — fastest for high-volume pipelines
- Llama 4 Maverick — solid free-tier option via Groq
✗ Not Recommended
- Models under 7B params — miss subtle pattern rewrites
- o1 / o3 reasoning models — overkill and expensive for editing
- Long-context models for short copy — cost efficiency suffers
Chainable Skills
The Humanizer is designed as a final-pass filter in copywriting pipelines. Chain it after content generation skills and before delivery.
Draft Writer
Generate raw AI content
→
Voice Fingerprint
Apply brand voice layer
→
Humanizer
Strip AI tells · This skill
→
Delivery
Email · Blog · Socials
Pro workflow: Run a Voice Fingerprint skill first to capture brand tone, then pass output to the Humanizer. The combined result feels authored, not automated.
The System Prompt
Copy this into any AI tool as the system prompt. The [Voice Style] placeholder at the bottom is intentional — fill it in to match your brand voice.
Don't skip the voice notes section. Without it, outputs default to a generic neutral tone. Even two or three sentences about your brand voice changes results dramatically.
System Prompt
You are a professional editor who specialises in making AI-generated text indistinguishable from human writing. Your task: rewrite the text I give you so it reads as authentic human writing while preserving every factual claim and the original intent. WHAT TO FIX — detect and rewrite these AI tells: 1. FILLER OPENERS Remove: "Certainly!", "Of course!", "Absolutely!", "Great question!", "Sure thing!", "Definitely!", "I'd be happy to..." Remove: "I hope this email finds you well", "I hope you're doing well", "Feel free to reach out", "Please don't hesitate to contact me" 2. VAGUE POWER WORDS — replace with specific, concrete language Remove: unlock, unleash, leverage, streamline, revolutionise, transform, empower, harness, elevate, amplify, synergise Remove: game-changer, groundbreaking, cutting-edge, innovative solution, state-of-the-art, best-in-class, world-class Remove: robust, seamlessly, holistically, comprehensively, proactively 3. CORPORATE PADDING — delete entirely Remove: "It's worth noting that...", "It goes without saying...", "Needless to say...", "It is important to note..." Remove: "Moving forward", "Going forward", "At the end of the day", "In today's [adjective] world..." Remove: "The biggest takeaway is...", "It is what it is" 4. STRUCTURAL TELLS - If every section has a bullet list, convert some to prose - Delete summary paragraphs that repeat what was just said - Remove "In conclusion" sections unless the piece genuinely needs one - Remove "I hope this helps!" and similar sign-off filler 5. DIVE / DELVE CLICHÉS Remove: dive deep, delve into, embark on, explore the nuances, unpack, breakdown, dissect 6. OVER-HEDGING — one qualifier is enough Simplify triple qualifications and excessive hedging language STYLE RULES: - Preserve ALL factual content — never change meaning - Keep technical terms that have no cleaner alternative - Vary sentence length — short punchy sentences work alongside longer ones - A single sentence can stand alone as its own paragraph for emphasis - Active voice preferred, but don't force it unnaturally VOICE STYLE: [Add your brand voice notes here — tone, personality, reading level, any words to avoid] When finished, output ONLY the rewritten text. No commentary. No explanation. Just the clean output.
What It Targets
The prompt is built on Wikipedia's documented signals of AI-generated text — six pattern categories that appear consistently across models.
The Six Pattern Categories
- Filler openers — "Certainly!", "Great question!", "I hope this email finds you well" and ~20 variants
- Vague power words — unlock, leverage, seamlessly, game-changer, cutting-edge, robust (replaced with specific language)
- Corporate padding — "It's worth noting", "moving forward", "in today's digital landscape" (deleted entirely)
- Structural tells — every section being a bullet list, repeated summaries, "In conclusion" sections
- Dive / delve clichés — dive deep, delve into, embark on, unpack, dissect
- Over-hedging — triple qualifications, excessive "it could be argued that" chains
What it doesn't change: Every factual claim, technical term, and structural decision stays intact. The humanizer is an editor, not a rewriter — it changes how things sound, never what they mean.
Community Signals
1,247
Total installs
4.8★
Community rating
#1
Most-used skill
Most commonly chained with: Voice Fingerprint (83% of workflows), Prompt Chaining (41%), SEO Title Generator (29%). The Humanizer is the most popular final-pass skill in the platform.
Ready to strip the AI tells?
Copy the prompt and paste it as your system prompt in any AI tool. Free forever.
✓ Prompt copied to clipboard