AI Humanizer Benchmark

Leaderboard / Compare / UndetectedGPT vs WriteHuman

Head to head · September 2026 cycle

UndetectedGPT vs WriteHuman

UndetectedGPT finished 3.3 points ahead of WriteHuman in the September 2026 cycle, 84.90 to 81.60, ranking #1 against #3 in a field of 11. UndetectedGPT scored higher on 3 of 7 detectors and 5 of 7 writing categories. WriteHuman came out ahead on Bypass rate (90.4 vs 86.3), Copyleaks (63.6 vs 21.2), Winston AI (82.1 vs 54.4). The bypass-rate gap, 90.4 to 86.3, sits inside both tools' 95% confidence intervals, so treat bypass as a draw.

UndetectedGPTLeads
undetectedgpt.ai
84.90/100
Overall score
#1/11
Rank
+3.3
UndetectedGPT leads
Detectors
34
Categories
52
Components
41
81.60/100
Overall score
#3/11
Rank
Last tested
2026-09-03
Prompts
33
Methodology
v1.0.0

Score components

What the overall score is made of: bypass rate weighs 42%, meaning preservation 32%, readability 16%, consistency across categories 10%. Output-quality penalties come off the total. How scoring works

UndetectedGPTWriteHuman
Bypass rateCIs overlapWriteHuman by 4.1
86.3
90.4
Meaning preservationUndetectedGPT by 8.6
84.8
76.2
ReadabilityUndetectedGPT by 5.0
77.3
72.4
ConsistencyUndetectedGPT by 5.0
91.5
86.4
Penaltiesfewer is betterUndetectedGPT by 1.0
None
−1.0

CIs overlap marks a gap that sits inside both tools' 95% confidence intervals for bypass rate; it may not survive another cycle.

Detector by detector

Share of each tool's outputs that a detector classified as human-written. WriteHuman took 4 of 7.

UndetectedGPTWriteHuman
GPTZeroUndetectedGPT by 23.5
80.0
56.5
Winston AIWriteHuman by 27.7
54.4
82.1
Originality.aiUndetectedGPT by 20.0
79.2
59.2
ZeroGPTWriteHuman by 6.9
85.1
92.0
GrammarlyWriteHuman by 4.8
89.3
94.2
QuillBotUndetectedGPT by 0.4
92.7
92.2
CopyleaksWriteHuman by 42.4
21.2
63.6

Writing categories

Category score per writing context: bypass credit only for real rewrites, blended with the category's own meaning preservation and readability. Small categories swing; the prompt count is shown on each row.

Academic essayApplication essayBlog postBusiness emailMarketing copyDiscussion boardNews article
UndetectedGPTWriteHumanOuter ring = 100
UndetectedGPTWriteHuman
Academic essay6 promptsUndetectedGPT by 3.0
89.2
86.2
Application essay6 promptsUndetectedGPT by 1.3
86.2
84.9
Blog post6 promptsWriteHuman by 3.0
77.6
80.6
Business email3 promptsUndetectedGPT by 10.2
79.7
69.5
Marketing copy6 promptsWriteHuman by 0.4
85.8
86.2
Discussion board3 promptsUndetectedGPT by 7.2
90.3
83.1
News article3 promptsUndetectedGPT by 6.3
85.9
79.6

Where each one wins

Every measure above that one tool won outright, largest margin first within each group.

UndetectedGPT

12 wins
  • Meaning preservation Score84.8 vs 76.2
  • Consistency Score91.5 vs 86.4
  • Readability Score77.3 vs 72.4
  • Penalties ScoreNone vs −1.0
  • GPTZero Detector80.0 vs 56.5
  • Originality.ai Detector79.2 vs 59.2
  • QuillBot Detector92.7 vs 92.2
  • Business email Category79.7 vs 69.5
  • Discussion board Category90.3 vs 83.1
  • News article Category85.9 vs 79.6
  • Academic essay Category89.2 vs 86.2
  • Application essay Category86.2 vs 84.9

WriteHuman

7 wins
  • Bypass rate Score90.4 vs 86.3
  • Copyleaks Detector63.6 vs 21.2
  • Winston AI Detector82.1 vs 54.4
  • ZeroGPT Detector92.0 vs 85.1
  • Grammarly Detector94.2 vs 89.3
  • Blog post Category80.6 vs 77.6
  • Marketing copy Category86.2 vs 85.8

Questions people ask

Is UndetectedGPT better than WriteHuman?

UndetectedGPT finished 3.3 points ahead of WriteHuman in the September 2026 cycle, 84.90 to 81.60, ranking #1 against #3 in a field of 11. UndetectedGPT scored higher on 3 of 7 detectors and 5 of 7 writing categories. Both tools ran the same 33 prompts, scored by the same 7 detectors, under methodology v1.0.0; every input, output, and verdict is published.

Which bypasses AI detectors better, UndetectedGPT or WriteHuman?

WriteHuman posted the higher bypass rate, 90.4 vs 86.3, and scored higher on 4 of 7 detectors. UndetectedGPT led on GPTZero, Originality.ai and QuillBot; WriteHuman led on Winston AI, ZeroGPT, Grammarly and Copyleaks. The overall bypass gap is inside both 95% confidence intervals, so it is not a reliable difference.

Do UndetectedGPT and WriteHuman keep the original meaning?

UndetectedGPT preserved meaning better, 84.8 vs 76.2 on our 0–100 similarity scale. On readability, UndetectedGPT rated higher, 77.3 vs 72.4. Penalties this cycle: UndetectedGPT None, WriteHuman −1.0.

More matchups

Add a third tool to this comparison·Every matchup

Go deeper