Bypass ALL major AI detectors — GPTZero, Turnitin (AIW-2 + AIR-1 + anti-humanizer), Originality.ai, QuillBot, ZeroGPT, Copyleaks, and every university-grade detector worldwide.
Built by Mohammad Sahil | GitHub: imsv1301
-
Detect — Paste text, click Detect. Gets AI probability score (0-100%), verdict (HUMAN/MIXED/AI), and specific AI tells found across 25 detection patterns.
-
Humanize — Click Humanize. Runs a 3-pass pipeline that rewrites text to bypass every detector:
- Pass 1: Deep restructuring (not paraphrasing) with perplexity injection + burstiness variation
- Pass 2: Silent audit finds every remaining AI tell
- Pass 3: Final revision fixes all remaining tells
-
Results — Shows changes summary + estimated detector bypass rate.
This isn't a simple synonym swapper or paraphraser. It targets the actual detection mechanisms:
| Detection signal | How UnMask.AI defeats it |
|---|---|
| Perplexity (word predictability) | Injects unexpected but natural word choices that push PPL above human/AI threshold |
| Burstiness (sentence variation) | Aggressively mixes 3-word fragments with 40+ word sentences. No 3 consecutive similar-length sentences allowed |
| Turnitin AIR-1 (paraphrase detection) | Genuinely restructures argument flow and paragraph logic, not just rewording |
| Turnitin anti-humanizer (Aug 2025) | Avoids uniform rewriting patterns that humanizer tools produce |
| QuillBot 10-pattern classifier | Eliminates all 10 flagged patterns: formal tone, predictable structure, generic language, etc. |
| 30+ banned AI vocabulary words | Removes delve, leverage, tapestry, nuanced, foster, robust, pivotal, transformative, etc. |
| Structural tells | Removes em dashes, rule-of-three, -ing phrases, copula avoidance, synonym cycling |
https://imsv1301.github.io/unmask-ai/
- Visit the live demo link above
- Enter your Anthropic API key
- Paste text, click Detect or Humanize
- Download
index.html - Open it in any browser
- Enter your API key and use it
No server, no installation, no dependencies. It's a single HTML file.
- Go to console.anthropic.com
- Sign up with Google or email (free to create account)
- Click "API Keys" in the left sidebar
- Click "Create Key" → copy it
- Go to Billing → add $5 minimum credit
- Paste the key into UnMask.AI
| What | Cost |
|---|---|
| Account creation | Free |
| Minimum credit | $5 (one time) |
| Per 1000 words humanized | ~$0.07 (₹6) |
| $5 gets you | ~70 humanizations |
That's enough for an entire semester of essays.
- Pass 1 (rewrite): ~1500 input + ~1500 output tokens
- Pass 2 (audit): ~2000 input + ~500 output tokens
- Pass 3 (final fix): ~3500 input + ~1500 output tokens
- Total: ~7000 input + ~3500 output tokens = ~$0.07
- Single HTML file (no framework, no build step)
- Anthropic Claude Sonnet 4 API with streaming
- Space Mono + Syne fonts
- Dark theme (#0a0a0f background)
- Purple/cyan accent palette
- GPTZero (7-component system)
- Turnitin (AIW-2 + AIR-1 + anti-humanizer 2025)
- Originality.ai
- QuillBot AI Detector
- ZeroGPT
- Copyleaks
- iThenticate
- Compilatio
- PlagScan / Urkund
- Any detector using perplexity + burstiness analysis
Send them this message:
Free AI text humanizer that actually bypasses Turnitin + GPTZero + all detectors:
- Go to https://imsv1301.github.io/unmask-ai/
- Get an API key from console.anthropic.com ($5 credit = 70+ uses)
- Paste text → click Humanize → done
Costs ₹6 per 1000 words. Works against every university detector.
MIT — use it however you want.
Built with research into how GPTZero, Turnitin, Originality.ai, QuillBot, and ZeroGPT actually detect AI text at the algorithmic level.