November 2026 · 750 outputs · 15 services
Originality.ai 5.0 detector benchmark
Model: model-5.0. Weight in the humanization composite: 20%.
Native prediction_label=human; mixed and ai fail. Displayed human score is 100 minus AI-assisted percentage; the native label determines pass.
Retains the 20% brand budget. Turbo was removed and Max was not run; no duplicate vendor model weight. Reference. None of these weights is a measured current detector accuracy.
| Humanizer | Passes/50 | Pass rate | Human-score average |
|---|---|---|---|
| Phrasly UltraModel: Ultra · Mode: Max | 37/50 | 74.00% | 75.20 |
| StealthGPT SuperModel: Super · Mode: Quality; rephrase enabled | 19/50 | 38.00% | 44.38 |
| Undetectable.aiModel: v11sr · Mode: University; General Writing; More Human | 1/50 | 2.00% | 14.28 |
| WriteHumanModel: Enhanced | 0/50 | 0.00% | 13.42 |
| RephrasyModel: Superstealth | 0/50 | 0.00% | 24.94 |
| Clever HumanizerModel: Not exposed · Mode: Normal; automatic Best Pick | 0/50 | 0.00% | 19.52 |
| ZeroGPT HumanizerModel: v11sr · Mode: University; General Writing; More Human | 0/50 | 0.00% | 13.36 |
| BypassGPTModel: Enhanced | 0/50 | 0.00% | 3.62 |
| GrubbyModel: Not exposed · Mode: Academic | 0/50 | 0.00% | 7.42 |
| Grammarly ProModel: Not exposed · Mode: The Naturalist; Humanizer rewrites only · Plan: Pro | 0/50 | 0.00% | 5.64 |
| HumbotModel: Advanced | 0/50 | 0.00% | 3.60 |
| SolvelyModel: Not exposed · Mode: Humanize; Original result | 0/50 | 0.00% | 3.98 |
| RewriteAIModel: Not exposed · Mode: Single available API model | 0/50 | 0.00% | 3.38 |
| QuillBotModel: Not exposed · Mode: Standard · Plan: Premium | 0/50 | 0.00% | 5.12 |
| HIX BypassModel: Not exposed · Mode: Latest | 0/50 | 0.00% | 5.10 |
Vendor verdicts
Native labels, shown as counts of 50 texts per humanizer. Mixed fails the strict-human pass rule; it is kept separate from AI.
| Humanizer | human (count) | mixed (count) | ai (count) |
|---|---|---|---|
| Phrasly Ultra | 37 | 3 | 10 |
| StealthGPT Super | 19 | 5 | 26 |
| WriteHuman | 0 | 7 | 43 |
| Rephrasy | 0 | 30 | 20 |
| Undetectable.ai | 1 | 6 | 43 |
| Clever Humanizer | 0 | 18 | 32 |
| ZeroGPT Humanizer | 0 | 9 | 41 |
| BypassGPT | 0 | 0 | 50 |
| Grubby | 0 | 3 | 47 |
| Grammarly Pro | 0 | 0 | 50 |
| Humbot | 0 | 0 | 50 |
| Solvely | 0 | 0 | 50 |
| RewriteAI | 0 | 0 | 50 |
| QuillBot | 0 | 0 | 50 |
| HIX Bypass | 0 | 0 | 50 |
This measures bypass of the tested detector, not its false-positive rate or accuracy on independently verified human writing. A high pass rate alone cannot establish that the detector is bad.
Per-output scores CSV · Frozen inputs and prompts CSV · Scoring manifest · Verification bundle · GitHub report · Earlier editions