HumanizerEval

Leaderboard / Compare / Rephrasy vs Undetectable.ai

September 2026 review · same 50 texts

Rephrasy vs Undetectable.ai

In the September 2026 HumanizerEval review, Rephrasy and Undetectable.ai rewrote the same 50 English texts. Rephrasy cleared all 5 detectors on 20 of them; Undetectable.ai on 15. The difference is mostly Pangram: on 12 samples Rephrasy passed it where Undetectable.ai failed, and the reverse happened on 5. On ZeroGPT and Winston AI the two were close to even.

Per-sample agreement

Per-sample agreement, September 2026. Counts are out of the 50 shared samples, computed at the 50% human-score threshold. Margin counts the samples only one tool passed, higher count first.
DetectorPassed more oftenMarginBoth passedNeither passed
Pangram 4.0Rephrasy12 to 51122
GPTZero 2026-09-13-baseUndetectable.ai4 to 0442
Originality.ai turboUndetectable.ai6 to 2411
ZeroGPT defaultTied0 to 0500
Winston AI 5.0Rephrasy1 to 0490
All 5 detectorsRephrasy9 to 41126

Every count is computed from scores.csv at build time, not transcribed. “Neither passed” counts the texts that defeated both tools; on Pangram that was 22 of 50 here.

Quality and output discipline

Quality and output discipline, September 2026.
MeasureRephrasyUndetectable.ai
36%26%
30%28%
92%94%
60%78%
1.131.21
37
00
v4University / General Writing / More Human, v11sr
Websiterephrasy.ai undetectable.ai

On per-sample quality scores, Rephrasy scored higher on 18 samples, Undetectable.ai on 18, and the two tied on 14. Rephrasy’s output was shorter on 33 of 50 samples; mean length ratios were 1.13 and 1.21, or 13% longer than the input and 21% longer than the input respectively.

Where each one outperforms the other

Rephrasy

  • Pangram: passed 23 of 50 against 16, including 12 samples where Undetectable.ai failed, against 5 the other way.
  • Originality.ai: passed 2 samples where Undetectable.ai failed, even though it was detected more often overall, passing 43 to 47.
  • Winston AI: passed 50 of 50 against 49, including 1 sample where Undetectable.ai failed, against 0 the other way.
  • Factual consistency: 36% against 26%.
  • Naturalness: 30% against 28%.
  • Length: shorter output on 33 of 50 samples, which matters under a word limit.

Undetectable.ai

  • Pangram: passed 5 samples where Rephrasy failed, even though it was detected more often overall, passing 16 to 23.
  • GPTZero: passed 48 of 50 against 44, including 4 samples where Rephrasy failed, against 0 the other way.
  • Originality.ai: passed 47 of 50 against 43, including 6 samples where Rephrasy failed, against 2 the other way.
  • ZeroGPT: tied at 50 of 50 but on a higher mean human score, 95.0% to 92.7%. It passed 0 samples where Rephrasy failed.
  • Syntax integrity: 94% against 92%.
  • Stance preservation: 78% against 60%.

Rephrasy vs Undetectable.ai FAQ

Is Rephrasy better than Undetectable.ai?

In the September 2026 HumanizerEval review, Rephrasy cleared all 5 detectors on 20 of the 50 shared samples and Undetectable.ai on 15. On ZeroGPT and Winston AI the two were within 6 points of each other. Quality ran 18 samples to 18, with 14 ties.

Which humanizer is best for Pangram, Rephrasy or Undetectable.ai?

Rephrasy. In September 2026, Rephrasy passed Pangram on 23 of the 50 shared samples and Undetectable.ai on 16. On 12 samples Rephrasy passed where Undetectable.ai failed; the reverse happened on 5. Both failed on 22.

Which humanizer changes the text less?

Neither. The two scored higher on per-sample quality equally often, 18 samples each. Rephrasy’s outputs averaged 1.13× the input length and passed factual consistency on 36%; Undetectable.ai’s averaged 1.21× and passed on 26%. On stance preservation the split was 60% to 78%. Rephrasy produced the shorter output on 33 of 50 samples.

More comparisons