Skip to content
The September 2026 results are in: WriteHuman holds #1, and newcomer SupWriter debuts at #5.Read the full analysis →
HumanizerBench

Head to head · September 2026 cycle

Phrasly vs StealthGPT

Phrasly finished 2.5 points ahead of StealthGPT in the September 2026 cycle, 67.58 to 65.10, ranking #8 against #9 in a field of 14. Phrasly scored higher on 1 of 5 detectors and 4 of 7 writing categories. StealthGPT came out ahead on academic essay prompts (79.4 vs 64.5), marketing copy prompts (73.1 vs 61.5) and penalties (−5.0 vs −6.0). The bypass-rate gap, 82.1 to 79.6, sits inside both tools' 95% confidence intervals, so treat bypass as a draw.

67.58 /100
overall score
#8 /14
rank ▼ 1
+2.48
← Phrasly leads
Detectors
1 3 (1 tie)
Categories
4 3
Components
3 2
65.10 /100
overall score
#9 /14
rank ▲ 3
Last tested
Prompts
33
Methodology
v1.2.0

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 →

Bypass rate CIs overlap Phrasly by 2.5
82.1 79.6
Meaning preservation CIs overlap Phrasly by 6.7
66.1 59.4
Readability StealthGPT by 5.6
65.4 70.9
Consistency Phrasly by 12.0
75.0 63.1
Penalties fewer is better StealthGPT by 1.0
−6.0 −5.0

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

Detector by detector

Share of each tool's outputs that a detector classified as human-written. StealthGPT took 3 of 5, with 1 tied.

GPTZero Phrasly by 4.0
81.4 77.4
Originality.ai StealthGPT by 6.1
83.4 89.5
Copyleaks Even
81.8 81.8
Winston AI StealthGPT by 5.1
61.6 66.6
ZeroGPT StealthGPT by 0.9
68.0 68.9

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. How category scores work →

Academic Essay Application Essay Blog Post Business Email Marketing Copy Discussion Board News Article Academic Essay: Phrasly 64.5 · StealthGPT 79.4 Application Essay: Phrasly 78.1 · StealthGPT 60.3 Blog Post: Phrasly 57.7 · StealthGPT 60.7 Business Email: Phrasly 51.7 · StealthGPT 42.7 Marketing Copy: Phrasly 61.5 · StealthGPT 73.1 Discussion Board: Phrasly 60.6 · StealthGPT 60.1 News Article: Phrasly 78.3 · StealthGPT 64.9
Phrasly StealthGPT Outer ring = 100
Academic Essay 6 prompts StealthGPT by 14.9
64.5 79.4
Application Essay 6 prompts Phrasly by 17.8
78.1 60.3
Blog Post 6 prompts StealthGPT by 3.0
57.7 60.7
Business Email 3 prompts Phrasly by 9.0
51.7 42.7
Marketing Copy 6 prompts StealthGPT by 11.6
61.5 73.1
Discussion Board 3 prompts Phrasly by 0.4
60.6 60.1
News Article 3 prompts Phrasly by 13.4
78.3 64.9

Head-to-head history

Overall score in every cycle both tools were tested. They have finished ahead 2 times each across 4 shared cycles.

40 50 60 70 80 June 2026 July 2026 August 2026 September 2026 June 2026: StealthGPT 71.47 (#7) June 2026: Phrasly 71.10 (#8) July 2026: StealthGPT 61.52 (#8) July 2026: Phrasly 61.47 (#9) August 2026: StealthGPT 45.55 (#12) August 2026: Phrasly 63.30 (#7) September 2026: StealthGPT 65.10 (#9) September 2026: Phrasly 67.58 (#8) 67.6 65.1
Phrasly StealthGPT
Cycle Phrasly StealthGPT Margin
September 2026 67.58 #8 65.10 #9 ← 2.5
August 2026 63.30 #7 45.55 #12 ← 17.8
July 2026 61.47 #9 61.52 #8 → 0.1
June 2026 71.10 #8 71.47 #7 → 0.4

Where each one wins

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

Phrasly

8 wins
  • Consistency Score 75.0 vs 63.1
  • Meaning preservation Score 66.1 vs 59.4
  • Bypass rate Score 82.1 vs 79.6
  • GPTZero Detector 81.4 vs 77.4
  • Application Essay Category 78.1 vs 60.3
  • News Article Category 78.3 vs 64.9
  • Business Email Category 51.7 vs 42.7
  • Discussion Board Category 60.6 vs 60.1

StealthGPT

8 wins
  • Penalties Score −5.0 vs −6.0
  • Readability Score 70.9 vs 65.4
  • Originality.ai Detector 89.5 vs 83.4
  • Winston AI Detector 66.6 vs 61.6
  • ZeroGPT Detector 68.9 vs 68.0
  • Academic Essay Category 79.4 vs 64.5
  • Marketing Copy Category 73.1 vs 61.5
  • Blog Post Category 60.7 vs 57.7

Questions people ask

Is Phrasly better than StealthGPT?

Phrasly finished 2.5 points ahead of StealthGPT in the September 2026 cycle, 67.58 to 65.10, ranking #8 against #9 in a field of 14. Phrasly scored higher on 1 of 5 detectors and 4 of 7 writing categories. Both tools ran the same 33 prompts, scored by the same 5 detectors, under methodology v1.2.0; every input, output and verdict is published.

Which bypasses AI detectors better, Phrasly or StealthGPT?

Phrasly posted the higher bypass rate, 82.1 vs 79.6, and scored higher on 1 of 5 detectors. Phrasly led on GPTZero; StealthGPT led on Originality.ai, Winston AI and ZeroGPT. The overall bypass gap is inside both 95% confidence intervals, so it is not a reliable difference.

Do Phrasly and StealthGPT keep the original meaning?

Phrasly preserved meaning better, 66.1 vs 59.4 on our 0–100 similarity scale, though the confidence intervals overlap. On readability, StealthGPT rated higher, 70.9 vs 65.4. Penalties this cycle: Phrasly −6.0, StealthGPT −5.0.