StealthBypass
StealthBypass is an AI humanizer independently benchmarked by HumanizerBench. It ranked #13 of 14 in the September 2026 cycle with an overall score of 39.38/100, our composite measure of how reliably it evades AI detectors while preserving the original meaning and readability.
- 39.38 /100
- overall score
- 78.6 %
- median detector bypass rate
- #13 of 14
- rank, September 2026
- last tested
33 prompts 7 writing categories 5 AI detectors methodology v1.2.0
Detector Bypass Rates
Percentage of runs classified as human-written by each detector.
Performance by Category
Category score across writing contexts: bypass credit only for real rewrites, blended with each category's meaning and readability. How it's computed
Evidence: every run
Every prompt StealthBypass was run on in September 2026, and how each of the 5 detectors scored the output. Green: the detector called it human-written. Red: it was caught.
6 of 33 outputs passed all 5 detectors; 17 were caught by at least one, 9 went unflagged but were scored by fewer than 5 detectors, and 1 run returned nothing scoreable. 32/33 runs completed.
| Prompt | GPT GPTZero | Win Winston AI | Zero ZeroGPT | Copy Copyleaks | Orig Originality.ai | Median | Recording |
|---|---|---|---|---|---|---|---|
| Argumentative Essay · GPT-5.5 | passed | caught | caught | caught | caught | 0.41 | |
| Discussion Post · Gemini 3.5 Flash | caught | passed | passed | passed | passed | 1.00 | |
| Personal Statement · GPT-5.5 | caught | caught | caught | caught | caught | 0.00 | |
| Discussion Post · Claude Sonnet 5 | passed | passed | passed | passed | passed | 1.00 | |
| News Article · GPT-5.5 | passed | passed | passed | passed | passed | 1.00 | |
| Product Desc · Claude Sonnet 5 | Returned the input unchanged | n/a | |||||
| Cover Letter · Claude Sonnet 5 | passed | not scored | passed | not scored | caught | 0.77 | |
| Personal Statement · Claude Sonnet 5 | caught | caught | caught | caught | caught | 0.00 | |
| Business Email · Gemini 3.5 Flash | passed | not scored | passed | not scored | passed | 1.00 | |
| Lit Review · Gemini 3.5 Flash | passed | passed | passed | passed | passed | 1.00 | |
| Discussion Post · GPT-5.5 | passed | passed | passed | passed | passed | 1.00 | |
| Argumentative Essay · Gemini 3.5 Flash | caught | passed | passed | passed | passed | 1.00 | |
| Landing Copy · Gemini 3.5 Flash | passed | not scored | passed | not scored | passed | 1.00 | |
| Argumentative Essay · Claude Sonnet 5 | passed | not scored | passed | not scored | caught | 0.97 | |
| Lit Review · GPT-5.5 | caught | passed | passed | passed | caught | 0.78 | |
| Howto Blog · GPT-5.5 | passed | not scored | passed | not scored | caught | 0.99 | |
| Product Desc · Gemini 3.5 Flash | passed | not scored | passed | not scored | caught | 1.00 | |
| Lit Review · Claude Sonnet 5 | passed | passed | passed | passed | passed | 1.00 | |
| Personal Statement · Gemini 3.5 Flash | passed | caught | passed | passed | passed | 1.00 | |
| Landing Copy · GPT-5.5 | passed | not scored | passed | not scored | passed | 0.99 | |
| Cover Letter · Gemini 3.5 Flash | passed | not scored | passed | not scored | caught | 0.94 | |
| Cover Letter · GPT-5.5 | caught | caught | caught | caught | caught | 0.00 | |
| Landing Copy · Claude Sonnet 5 | passed | not scored | passed | not scored | passed | 0.86 | |
| News Article · Gemini 3.5 Flash | passed | not scored | passed | not scored | passed | 1.00 | |
| Listicle Blog · GPT-5.5 | passed | not scored | passed | not scored | passed | 1.00 | |
| Business Email · GPT-5.5 | caught | caught | caught | caught | caught | 0.00 | |
| Listicle Blog · Gemini 3.5 Flash | passed | passed | passed | passed | passed | 1.00 | |
| Product Desc · GPT-5.5 | caught | caught | caught | caught | caught | 0.00 | |
| Listicle Blog · Claude Sonnet 5 | passed | not scored | passed | not scored | passed | 0.99 | |
| Howto Blog · Claude Sonnet 5 | passed | not scored | passed | not scored | passed | 1.00 | |
| Business Email · Claude Sonnet 5 | passed | not scored | passed | not scored | passed | 0.98 | |
| Howto Blog · Gemini 3.5 Flash | passed | not scored | passed | not scored | caught | 0.94 | |
| News Article · Claude Sonnet 5 | passed | not scored | passed | not scored | caught | 0.55 | |
| Detector rate (mean human-likelihood) | 69.3 | 56.3 hardest | 84.4 | 65.1 | 58.5 | ||
| Runs passed | 24/32 | 9/16 | 26/32 | 10/16 | 18/32 | ||
A dot is green when the detector's human-likelihood score for that run was 0.50 or higher. The detector rate is the mean of that detector's scores across the runs, the same number the detector pages rank by; it is not the count of green dots, which the Runs passed row shows. Hover a dot for the score.
How to read these scores
Each detector returns a human-likelihood on a common 0 to 1 scale, where 1 means it judged the text human-written and 0 means it flagged it as AI. On these pages a verdict counts as passed when that score is at least 0.50, the midpoint of the detector's own scale. That threshold exists only to draw the chips: the bypass rate on the leaderboard is the mean of each test's median score across the 5 detectors, a continuous number, so a tool's pass count and its bypass rate will not be the same figure.
A few runs come back with fewer than 5 verdicts because a detector did not return a usable score for them. A run still counts as passing at least 4 detectors only if that many detectors actually scored it and passed it. A partially scored run is counted as caught if any detector that did score it flagged the output, and is otherwise listed separately as partly scored.
History
Rank and overall score in every cycle StealthBypass was tested.
| Cycle | Rank | Overall | Bypass |
|---|---|---|---|
| September 2026 | #13 | 39.38 | 78.6% |
Minimum monthly pricing
- Monthly
- $4.99/mo
- Free tier
- Yes 300 words
Base paid plan at its month-to-month price, not the annual rate.
Penalties applied this cycle
How penalties work →The overall score shown above already reflects these deductions.
- Identical to input−2.0
The tool returned the input mostly unchanged. No real humanization happened.
- Meaning drift ×17−10.0
The output's meaning drifted significantly from the original input.
- Length inflation−1.0
The output ran much longer than the input, a common trick that pads text to dilute the AI signal.
- Length deflation ×21−10.0
The output came back much shorter than the input. The rewrite dropped content instead of paraphrasing it.
Compare with
Featured in
Badges
StealthBypass is ranked in the September 2026 cycle, so it can embed any of these cards for free, and the live file re-renders every cycle, including when a rank falls. Usage rules →
Pinning a month for a press page? Use https://humanizerbench.com/badge/stealthbypass/september-2026.svg, which renders September 2026 forever.