October 2026 review · run October 22, 2026
The AI humanizer benchmark that tests against Pangram.
In the October 2026 review, StealthGPT has the best bypass rate among AI humanizers. It passed Pangram on 88% of texts, against 40% for the next best tool.
October 2026 AI humanizer leaderboard
| # | Humanizer | Pangram | GPTZero | ZeroGPT | Winston AI | ||||
|---|---|---|---|---|---|---|---|---|---|
| 1 | StealthGPT17/50 cleared all | 80.3% | 88% | 92% | 92% | 34% | 84% | 92% | 1.18 |
| 2 | Undetectable.ai1/50 cleared all | 71.3% | 40% | 86% | 98% | 4% | 100% | 100% | 1.27 |
| 3 | WriteHuman0/50 cleared all | 69.7% | 26% | 90% | 100% | 6% | 96% | 100% | 1.06 |
| 4 | Rephrasy0/50 cleared all | 62.3% | 26% | 70% | 86% | 0% | 92% | 100% | 1.22 |
| 5 | Grubby0/50 cleared all | 62.3% | 10% | 68% | 100% | 0% | 96% | 100% | 1.19 |
| 6 | RewriteAI0/50 cleared all | 56.3% | 2% | 54% | 98% | 0% | 90% | 94% | 1.21 |
| 7 | Phrasly0/50 cleared all | 48.7% | 4% | 44% | 72% | 0% | 82% | 90% | 1.10 |
| 8 | HIX Bypass0/50 cleared all | 44.0% | 0% | 6% | 66% | 0% | 92% | 100% | 1.15 |
| 9 | Humbot0/50 cleared all | 44.0% | 0% | 0% | 76% | 0% | 88% | 100% | 1.15 |
| 10 | BypassGPT0/50 cleared all | 41.3% | 0% | 4% | 56% | 0% | 88% | 100% | 1.15 |
| 11 | Ryne0/50 cleared all | 0.7% | 0% | 0% | 0% | 0% | 4% | 0% | 1.02 |
| Source textNo humanizer | 0.3% | 0% | 0% | 0% | 0% | 2% | 0% | — |
Each column counts how many of the 50 texts that detector called human-written. Rank is the average of the six columns. The last row is the untouched source text, scored by the same detectors. Full methodology.
Bypass by detector, and bypass against quality
The first chart shows each tool’s strict-bypass rate on every detector. The second plots mean bypass against quality pass rate.
- Pangram
- GPTZero
- Originality.ai Turbo
- Originality.ai 5.0
- ZeroGPT
- Winston AI
Performance by detector
Undetectability score is model-dependent, so each detector gets its own page and its own ranking. A tool that bypasses one detector can be detected on another.
Best AI humanizer for Pangram
88%
StealthGPT
Best AI humanizer for GPTZero
92%
StealthGPT (tied)
Best AI humanizer for Originality.ai 5.0
34%
StealthGPT
Best AI humanizer for ZeroGPT
100%
Undetectable.ai
Best AI humanizer for Winston AI
100%
StealthGPT (tied)
Originality.ai 5.0 is the hardest detector in the panel
In October 2026, StealthGPT passed Originality.ai 5.0 model-5.0 on 17 of 50 texts with a mean human score of 38.7%. Across the panel, the average tool passed Originality.ai 5.0 on 4% of texts, against 18% on Pangram, 47% on GPTZero, 77% on Originality.ai Turbo, 83% on ZeroGPT and 89% on Winston AI. Pangram, which neither of the two other public benchmarks tests, was passed by the average tool on 18% of the same texts.
Quality, reported separately
Bypassing AI detectors is worthless if the humanized text changes the meaning. Each output was checked with 4 factors in mind; a pass means no issue was found. These numbers never enter the ranking, so a weighting choice cannot decide the order.
| Humanizer | |||||
|---|---|---|---|---|---|
| Ryne | 87.5% | 64% | 92% | 100% | 94% |
| StealthGPT | 83.5% | 68% | 82% | 98% | 86% |
| Grubby | 79.5% | 76% | 64% | 100% | 78% |
| Phrasly | 75.5% | 50% | 70% | 98% | 84% |
| RewriteAI | 67.5% | 40% | 52% | 100% | 78% |
| WriteHuman | 59.0% | 38% | 50% | 98% | 50% |
| Rephrasy | 59.0% | 32% | 44% | 98% | 62% |
| Undetectable.ai | 52.5% | 32% | 28% | 90% | 60% |
| Humbot | 52.0% | 30% | 6% | 100% | 72% |
| HIX Bypass | 48.0% | 22% | 6% | 98% | 66% |
| BypassGPT | 47.5% | 24% | 0% | 100% | 66% |
In October 2026, Ryne had the highest mean quality pass rate at 87.5%, and BypassGPT the lowest at 47.5%.
Which humanizer for which job
- Best for students
- Pangram is used by Capella University, Substack, Quora, NewsGuard, Penguin Random House, and others. In this review StealthGPT led Pangram at 44 of 50.
- Best for marketers and SEO teams
- StealthGPT is the best for bypassing, including Google’s AI detection, which helps content rank higher. Google is less likely to rank pages it reads as machine-written.
- Best for publishers and writers
- StealthGPT keeps a human touch in the writing, which is what a byline needs. Its mean quality pass rate was 83.5%. Ryne had the highest quality pass rate, at 87.5%.
Recompute this page yourself
Every number above comes from one public CSV of per-sample detector scores. Clone the repo and recompute the aggregates:
git clone https://github.com/StealthGPT-Labs/humanizer-benchmark
cd humanizer-benchmark
python3 scripts/verify_report.py reports/2026-10-22-humanizer-comparisonscores.csv
50 rows of anonymized per-sample detector scores, output word counts and quality scores for all 11 tools. SHA-256 on the review archive.
Source texts
The 50 texts every tool received, with the prompt and model that wrote each one. Humanized outputs are not included.
leaderboard.json
The full ranking, per-detector results, quality pass rates and pairwise counts as machine-readable JSON.
Methodology v1.0
Detector versions, the strict-bypass definition, the sort rule, the settings rule, and the gaps we have not closed yet.
Why another humanizer benchmark
There are two other public humanizer benchmarks. One is run by WriteHuman and ranks WriteHuman first. One is run by UndetectedGPT and ranks UndetectedGPT first. They share a methodology template almost line for line, and neither tests Pangram, the one detector that reliably separates a strong humanizer from a weak one.
HumanizerEval is the first and only benchmark that tests against Pangram. It is the truest and most comprehensive benchmark available to give a full picture of humanizer performance.
Frequently asked questions
What is an AI humanizer?
An AI humanizer rewrites machine-generated text so that AI-content detectors are less likely to flag it. Most run a language model with a paraphrasing prompt plus post-processing that alters perplexity, burstiness and the other statistical patterns detectors look for, while trying to keep the original meaning intact. Whether that works depends almost entirely on which detector you face.
Which AI humanizer is best for Pangram?
In October 2026, StealthGPT passed Pangram v4 on 44 of 50 texts (88%). The next best was Undetectable.ai at 40%. See the Pangram page.
Why does Pangram give such different results from GPTZero?
Detectors use different models and different thresholds, and their human scores are not comparable. Pangram ships a classification head trained specifically to recognise humanizer output; GPTZero does not advertise one. On the same 50 texts in October 2026, the average tool passed GPTZero on 47% of texts and Pangram on 18%. That is why HumanizerEval reports every detector separately and never blends their scores.
Why should I trust these results?
Because you can recompute them. Every tool ran on the same 50 texts, through the same detector versions, with the settings recorded per tool. The per-sample scores are in a public CSV, and one command recomputes every number on this page.
Why is Turnitin not tested?
Turnitin offers no public API for this kind of testing, so it cannot be run at 50 samples per tool per review without a licensed institutional account. However, independent testing shows that StealthGPT Super bypasses Turnitin ~95% of the time.