HumanizerEval

Leaderboard / Compare / StealthGPT vs RewriteAI

September 2026 review · same 50 texts

StealthGPT Super vs RewriteAI

In the September 2026 HumanizerEval review, StealthGPT Super and RewriteAI rewrote the same 50 English texts. StealthGPT Super cleared all 5 detectors on 46 of them; RewriteAI on 13. The difference is mostly Pangram: on 32 samples StealthGPT passed it where RewriteAI failed, and the reverse never happened. On 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 4.0StealthGPT32 to 0144
GPTZero 2026-09-13-baseStealthGPT7 to 1420
Originality.ai turboStealthGPT5 to 0441
ZeroGPT defaultStealthGPT6 to 1430
Winston AI 5.0StealthGPT2 to 1470
All 5 detectorsStealthGPT33 to 0134

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 4 of 50 here.

Quality and output discipline

Quality and output discipline, September 2026.
MeasureStealthGPT SuperRewriteAI
68%54%
72%40%
98%100%
96%94%
1.041.09
02
00
Super, production outputAPI default, passages of up to 280 words (300-word API limit)
Websitestealthgpt.ai rewriteai.com

On per-sample quality scores, StealthGPT scored higher on 26 samples, RewriteAI on 5, and the two tied on 19. StealthGPT’s output was shorter on 29 of 50 samples; mean length ratios were 1.04 and 1.09, or 4% longer than the input and 9% longer than the input respectively.

Where StealthGPT outperforms RewriteAI

StealthGPT

  • Pangram: passed 46 of 50 samples, against 14 for RewriteAI.
  • GPTZero: passed 49 of 50 samples, against 43 for RewriteAI.
  • Originality.ai: passed 49 of 50 samples, against 44 for RewriteAI.
  • ZeroGPT: passed 49 of 50 samples, against 44 for RewriteAI.
  • Winston AI: passed 49 of 50 samples, against 48 for RewriteAI.
  • Factual consistency: kept facts accurate in 68% of outputs, against 54% for RewriteAI.
  • Naturalness: read fluently in 72% of outputs, against 40% for RewriteAI.
  • Stance preservation: preserved the source’s stance in 96% of outputs, against 94% for RewriteAI.
  • Length: shorter than RewriteAI on 29 of 50 samples, which matters under a word limit.

RewriteAI

  • Pangram: failed 36 of 50 samples, 32 of which StealthGPT passed.
  • GPTZero: failed 7 of 50 samples, 7 of which StealthGPT passed.
  • Originality.ai: failed 6 of 50 samples, 5 of which StealthGPT passed.
  • ZeroGPT: failed 6 of 50 samples, 6 of which StealthGPT passed.
  • Winston AI: failed 2 of 50 samples, 2 of which StealthGPT passed.
  • Factual consistency: introduced factual errors in 46% of outputs.
  • Naturalness: had awkward or clunky phrasing in 60% of outputs.
  • Stance preservation: distorted the source’s stance in 6% of outputs.
  • Length: longer than StealthGPT on 29 of 50 samples, so more to cut under a word limit.

StealthGPT vs RewriteAI FAQ

Is StealthGPT better than RewriteAI?

Yes. In the September 2026 HumanizerEval review, StealthGPT Super cleared all 5 detectors on 46 of the 50 shared samples and RewriteAI on 13. On Winston AI the two were within 6 points of each other. Quality ran 26 samples to 5, with 19 ties.

Which humanizer is best for Pangram, StealthGPT or RewriteAI?

StealthGPT Super. In September 2026, StealthGPT Super passed Pangram on 46 of the 50 shared samples and RewriteAI on 14. On 32 samples StealthGPT passed where RewriteAI failed; the reverse never happened. Both failed on 4.

Which humanizer changes the text less?

StealthGPT Super. It scored higher on per-sample quality on 26 of 50 samples, against 5. StealthGPT Super’s outputs averaged 1.04× the input length and passed factual consistency on 68%; RewriteAI’s averaged 1.09× and passed on 54%. On stance preservation the split was 96% to 94%. StealthGPT produced the shorter output on 29 of 50 samples.

More comparisons