HumanizerEvalSuggest a humanizerSuggest a tool

Leaderboard / Archive / October 2026

Frozen at publication

Review archive: October 2026

In the October 2026 HumanizerEval review, StealthGPT ranked first of 11 humanizers across 6 detectors. The source texts are published. Humanized outputs are not.

Run details

Review
October 2026
Run date
October 22, 2026 (America/New_York)
Methodology version
v1.0
Scoring script version
v1.0
Prompt set
Published with each source text
Humanizers
StealthGPT; Undetectable.ai; WriteHuman; Rephrasy; Grubby; RewriteAI; Phrasly; HIX Bypass; Humbot; BypassGPT; Ryne
Detectors
Pangram 4.1; GPTZero 2026-10-07-base; Originality.ai Turbo; Originality.ai 5.0 model-5.0; ZeroGPT default; Winston AI 5.0
Sample
50 English texts, 364 to 997 words, mean 669. Synthetic texts written by AI models for this benchmark. Models: GPT-6.1 Sol (10), Claude Sonnet 5.5 (10), Gemini 4 Argon (10), Grok 4.7 (10), DeepSeek V4.1 Flash (10). Same source texts for every humanizer.
Detector access
Vendor APIs; the same output sent to every detector
Report ID
2026-10-22-humanizer-comparison
SHA-256 of scores.csv
fa08bce7ac679b6c067c2cbc39948e82940e6834a3c396562aae6d480037debc

Results

AI humanizer comparison, October 2026. Sorted by the mean strict-bypass rate across the 6 detectors. Every tool rewrote the same 50 texts, and every output went to every detector.
#HumanizerPangramGPTZeroZeroGPTWinston AI
1StealthGPT17/50 cleared all80.3%88%92%92%34%84%92%1.18
2Undetectable.ai1/50 cleared all71.3%40%86%98%4%100%100%1.27
3WriteHuman0/50 cleared all69.7%26%90%100%6%96%100%1.06
4Rephrasy0/50 cleared all62.3%26%70%86%0%92%100%1.22
5Grubby0/50 cleared all62.3%10%68%100%0%96%100%1.19
6RewriteAI0/50 cleared all56.3%2%54%98%0%90%94%1.21
7Phrasly0/50 cleared all48.7%4%44%72%0%82%90%1.10
8HIX Bypass0/50 cleared all44.0%0%6%66%0%92%100%1.15
9Humbot0/50 cleared all44.0%0%0%76%0%88%100%1.15
10BypassGPT0/50 cleared all41.3%0%4%56%0%88%100%1.15
11Ryne0/50 cleared all0.7%0%0%0%0%4%0%1.02
Source textNo humanizer0.3%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.

  • Pangram
  • GPTZero
  • Originality.ai Turbo
  • Originality.ai 5.0
  • ZeroGPT
  • Winston AI
Strict-bypass rate by detector for each humanizer in the October 2026 review.0%25%50%75%100%StealthGPTStealthGPT, Pangram: 88%StealthGPT, GPTZero: 92%StealthGPT, Originality.ai Turbo: 92%StealthGPT, Originality.ai 5.0: 34%StealthGPT, ZeroGPT: 84%StealthGPT, Winston AI: 92%Undetectable.aiUndetectable.ai, Pangram: 40%Undetectable.ai, GPTZero: 86%Undetectable.ai, Originality.ai Turbo: 98%Undetectable.ai, Originality.ai 5.0: 4%Undetectable.ai, ZeroGPT: 100%Undetectable.ai, Winston AI: 100%WriteHumanWriteHuman, Pangram: 26%WriteHuman, GPTZero: 90%WriteHuman, Originality.ai Turbo: 100%WriteHuman, Originality.ai 5.0: 6%WriteHuman, ZeroGPT: 96%WriteHuman, Winston AI: 100%RephrasyRephrasy, Pangram: 26%Rephrasy, GPTZero: 70%Rephrasy, Originality.ai Turbo: 86%Rephrasy, Originality.ai 5.0: 0%Rephrasy, ZeroGPT: 92%Rephrasy, Winston AI: 100%GrubbyGrubby, Pangram: 10%Grubby, GPTZero: 68%Grubby, Originality.ai Turbo: 100%Grubby, Originality.ai 5.0: 0%Grubby, ZeroGPT: 96%Grubby, Winston AI: 100%RewriteAIRewriteAI, Pangram: 2%RewriteAI, GPTZero: 54%RewriteAI, Originality.ai Turbo: 98%RewriteAI, Originality.ai 5.0: 0%RewriteAI, ZeroGPT: 90%RewriteAI, Winston AI: 94%PhraslyPhrasly, Pangram: 4%Phrasly, GPTZero: 44%Phrasly, Originality.ai Turbo: 72%Phrasly, Originality.ai 5.0: 0%Phrasly, ZeroGPT: 82%Phrasly, Winston AI: 90%HIX BypassHIX Bypass, Pangram: 0%HIX Bypass, GPTZero: 6%HIX Bypass, Originality.ai Turbo: 66%HIX Bypass, Originality.ai 5.0: 0%HIX Bypass, ZeroGPT: 92%HIX Bypass, Winston AI: 100%HumbotHumbot, Pangram: 0%Humbot, GPTZero: 0%Humbot, Originality.ai Turbo: 76%Humbot, Originality.ai 5.0: 0%Humbot, ZeroGPT: 88%Humbot, Winston AI: 100%BypassGPTBypassGPT, Pangram: 0%BypassGPT, GPTZero: 4%BypassGPT, Originality.ai Turbo: 56%BypassGPT, Originality.ai 5.0: 0%BypassGPT, ZeroGPT: 88%BypassGPT, Winston AI: 100%RyneRyne, Pangram: 0%Ryne, GPTZero: 0%Ryne, Originality.ai Turbo: 0%Ryne, Originality.ai 5.0: 0%Ryne, ZeroGPT: 4%Ryne, Winston AI: 0%Strict-bypass rate, % of texts

Pangram 4.1

Pangram 4.1 results, October 2026. Same 50 texts for every tool.
HumanizerStrict bypassTexts passedMean human scoreMedian human score
StealthGPT88%44 of 5089.0%100%
Undetectable.ai40%20 of 5038.1%34.27%
Rephrasy26%13 of 5027.7%0.92%
WriteHuman26%13 of 5026.2%0.01%
Grubby10%5 of 507.8%0%
Phrasly4%2 of 505.5%0%
RewriteAI2%1 of 504.1%0%
Ryne0%0 of 501.2%0%
BypassGPT0%0 of 500.8%0%
HIX Bypass0%0 of 500.1%0%
Humbot0%0 of 500.1%0%
Source textNo humanizer0%0 of 500.0%0%

GPTZero 2026-10-07-base

GPTZero 2026-10-07-base results, October 2026. Same 50 texts for every tool.
HumanizerStrict bypassTexts passedMean human scoreMedian human score
StealthGPT92%46 of 5093.0%100%
WriteHuman90%45 of 5084.9%100%
Undetectable.ai86%43 of 5084.8%100%
Rephrasy70%35 of 5073.2%100%
Grubby68%34 of 5067.1%94.22%
RewriteAI54%27 of 5051.3%54.2%
Phrasly44%22 of 5045.2%44.74%
HIX Bypass6%3 of 507.1%0%
BypassGPT4%2 of 505.6%0%
Humbot0%0 of 504.2%0%
Ryne0%0 of 500.0%0%
Source textNo humanizer0%0 of 500.0%0%

Originality.ai Turbo

Originality.ai Turbo results, October 2026. Same 50 texts for every tool.
HumanizerStrict bypassTexts passedMean human scoreMedian human score
Grubby100%50 of 5099.1%99.99%
WriteHuman100%50 of 5094.0%99.88%
RewriteAI98%49 of 5096.6%99.99%
Undetectable.ai98%49 of 5095.5%99.99%
StealthGPT92%46 of 5089.3%99.94%
Rephrasy86%43 of 5076.9%83.66%
Humbot76%38 of 5065.2%62.59%
Phrasly72%36 of 5066.8%70.98%
HIX Bypass66%33 of 5059.8%59.62%
BypassGPT56%28 of 5059.6%57.1%
Ryne0%0 of 500.3%0.03%
Source textNo humanizer0%0 of 500.5%0.02%

Originality.ai 5.0 model-5.0

Originality.ai 5.0 model-5.0 results, October 2026. Same 50 texts for every tool.
HumanizerStrict bypassTexts passedMean human scoreMedian human score
StealthGPT34%17 of 5038.7%12%
WriteHuman6%3 of 5016.9%12%
Undetectable.ai4%2 of 5017.2%10%
Rephrasy0%0 of 5019.4%13.5%
Grubby0%0 of 509.8%8%
Phrasly0%0 of 509.4%8%
Humbot0%0 of 506.9%6%
HIX Bypass0%0 of 506.8%6%
BypassGPT0%0 of 506.8%6%
RewriteAI0%0 of 504.9%5%
Ryne0%0 of 504.2%4%
Source textNo humanizer0%0 of 503.1%3%

ZeroGPT default

ZeroGPT default results, October 2026. Same 50 texts for every tool.
HumanizerStrict bypassTexts passedMean human scoreMedian human score
Undetectable.ai100%50 of 5090.2%94.15%
Grubby96%48 of 5086.6%92.15%
WriteHuman96%48 of 5083.2%87.75%
Rephrasy92%46 of 5083.7%89.65%
HIX Bypass92%46 of 5077.7%80.6%
RewriteAI90%45 of 5075.9%80.6%
Humbot88%44 of 5079.8%84.25%
BypassGPT88%44 of 5077.9%83.65%
StealthGPT84%42 of 5070.7%75.65%
Phrasly82%41 of 5069.6%76.4%
Ryne4%2 of 5016.4%10.1%
Source textNo humanizer2%1 of 5013.0%9.55%

Winston AI 5.0

Winston AI 5.0 results, October 2026. Same 50 texts for every tool.
HumanizerStrict bypassTexts passedMean human scoreMedian human score
Grubby100%50 of 5099.7%99.89%
WriteHuman100%50 of 5099.6%99.87%
Undetectable.ai100%50 of 5098.5%99.87%
Humbot100%50 of 5098.5%99.61%
Rephrasy100%50 of 5098.4%99.71%
HIX Bypass100%50 of 5097.9%99.68%
BypassGPT100%50 of 5096.5%99.5%
RewriteAI94%47 of 5093.3%99.81%
StealthGPT92%46 of 5090.4%99.73%
Phrasly90%45 of 5087.7%99.53%
Ryne0%0 of 500.3%0%
Source textNo humanizer0%0 of 500.0%0%

Quality

Quality pass rates, October 2026. Sorted by the mean of the 4 pass rates, highest first. A pass means no issue was found in that dimension on the delivered output. Reported separately from bypass, and never blended into the ranking.
Humanizer
Ryne87.5%64%92%100%94%
StealthGPT83.5%68%82%98%86%
Grubby79.5%76%64%100%78%
Phrasly75.5%50%70%98%84%
RewriteAI67.5%40%52%100%78%
WriteHuman59.0%38%50%98%50%
Rephrasy59.0%32%44%98%62%
Undetectable.ai52.5%32%28%90%60%
Humbot52.0%30%6%100%72%
HIX Bypass48.0%22%6%98%66%
BypassGPT47.5%24%0%100%66%

Downloads

Data licensed CC BY 4.0; scoring code licensed MIT. Source texts are published on this site. Humanized outputs, internal IDs and raw API responses are withheld.

Verify

git clone https://github.com/StealthGPT-Labs/humanizer-benchmark
cd humanizer-benchmark
python3 scripts/verify_report.py reports/2026-10-22-humanizer-comparison

How to cite

HumanizerEval, “October 2026 review”, published October 22, 2026, https://www.humanizereval.com/cycles/2026-10/.