Undetectable.ai
Undetectable.ai ranked #3 of 14 in the October 2026 cycle with an overall score of 71.18/100, after losing 5 points to penalties for length inflation. Winston AI passed its output most often (98.5%), and Pangram passed it least often (3.6%). It did best on application essays and worst on blog posts.
Undetectable.ai is a humanizer from Undetectable Inc., part of a suite that also includes an AI detector, essay writer, paraphraser and Chrome extension. The humanizer lets users pick a model and set a readability level and a purpose, such as essay or marketing material. A free version runs a basic model, and a free trial gives access to the other models.
Watch the 2-minute review of Undetectable.ai
- Overall score
- 71.18/100
- Detector bypass
- 74.6%
- Third place
- #3of 14
- Prompts tested
- 30
Detector Bypass Rates
Each detector's average human-likelihood score on this tool's outputs, where 100 means every output was scored as human.
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
Strengths
- Meaning preservation of 79.2, above the field median of 71.0
- Winston AI classified 98.5% of its runs as human-written
- Best writing category: application essays (76.4)
Weaknesses
- Lost 5 points to penalties for length inflation
- Pangram classified only 3.6% of its runs as human-written
- Weakest writing category: blog posts (58.1)
From the October 2026 cycle, compared with the other 13 tools ranked in that cycle.
Evidence: every run
Every prompt Undetectable.ai was run on in October 2026, and how each of the 6 detectors scored the output. Green: the detector called it human-written. Red: it was caught.
1 of 30 outputs passed all 6 detectors; 29 were caught by at least one.
| Prompt | GPT GPTZero | Pan Pangram | Win Winston AI | Zero ZeroGPT | Copy Copyleaks | Orig Originality.ai | Mean | Recording |
|---|---|---|---|---|---|---|---|---|
| Landing Copy · gemini-3-8-flash | passed | caught | passed | passed | caught | passed | 0.59 | |
| Product Desc · gemini-3-8-flash | caught | caught | passed | passed | passed | passed | 0.61 | |
| Landing Copy · gpt-6-sol | passed | caught | passed | caught | passed | caught | 0.48 | |
| Discussion Post · gpt-6-sol | passed | caught | passed | passed | passed | passed | 0.75 | |
| Product Desc · gpt-6-sol | passed | caught | passed | passed | passed | passed | 0.71 | |
| Lit Review · claude-sonnet-5-5 | passed | caught | passed | passed | passed | passed | 0.77 | |
| Lit Review · gpt-6-sol | passed | caught | passed | passed | passed | passed | 0.88 | |
| Cover Letter · gpt-6-sol | caught | caught | passed | passed | passed | passed | 0.65 | |
| Howto Blog · gemini-3-8-flash | passed | caught | passed | passed | passed | passed | 0.79 | |
| Lit Review · gemini-3-8-flash | passed | caught | passed | passed | passed | passed | 0.83 | |
| Landing Copy · claude-sonnet-5-5 | passed | caught | passed | passed | passed | caught | 0.66 | |
| Personal Statement · gemini-3-8-flash | passed | caught | passed | passed | passed | passed | 0.77 | |
| Argumentative Essay · gpt-6-sol | passed | caught | passed | passed | passed | passed | 0.78 | |
| Listicle Blog · gemini-3-8-flash | passed | caught | passed | passed | passed | passed | 0.81 | |
| Personal Statement · claude-sonnet-5-5 | passed | passed | passed | passed | passed | passed | 0.95 | |
| Howto Blog · gpt-6-sol | passed | caught | passed | passed | passed | passed | 0.75 | |
| Product Desc · claude-sonnet-5-5 | passed | caught | passed | passed | passed | passed | 0.77 | |
| Listicle Blog · claude-sonnet-5-5 | passed | caught | passed | passed | passed | passed | 0.77 | |
| Argumentative Essay · claude-sonnet-5-5 | passed | caught | passed | passed | caught | passed | 0.67 | |
| News Article · gpt-6-sol | passed | caught | passed | passed | passed | passed | 0.83 | |
| Howto Blog · claude-sonnet-5-5 | passed | caught | passed | caught | caught | passed | 0.53 | |
| News Article · claude-sonnet-5-5 | passed | caught | passed | passed | passed | passed | 0.83 | |
| Discussion Post · gemini-3-8-flash | caught | caught | passed | passed | passed | passed | 0.62 | |
| Discussion Post · claude-sonnet-5-5 | passed | caught | passed | passed | passed | passed | 0.79 | |
| Cover Letter · gemini-3-8-flash | passed | caught | passed | passed | passed | passed | 0.81 | |
| Listicle Blog · gpt-6-sol | passed | caught | passed | passed | passed | passed | 0.74 | |
| Cover Letter · claude-sonnet-5-5 | passed | caught | passed | passed | passed | passed | 0.82 | |
| Personal Statement · gpt-6-sol | passed | caught | passed | passed | passed | passed | 0.82 | |
| News Article · gemini-3-8-flash | passed | caught | passed | passed | passed | passed | 0.83 | |
| Argumentative Essay · gemini-3-8-flash | passed | caught | passed | passed | passed | passed | 0.78 | |
| Detector rate (mean human-likelihood) | 85.0 | 3.5 hardest | 98.5 | 83.8 | 87.9 | 88.7 | ||
| Runs passed | 27/30 | 1/30 | 30/30 | 28/30 | 27/30 | 28/30 |
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 scores from the 6 detectors, a continuous number, so a tool's pass count and its bypass rate will not be the same figure.
History
Overall score compared with other ranked tools.
5.2 points above the median of 13 other tools in October 2026.
The shaded band is the 25th–75th percentile of other tools in each cycle. Scores are out of 100; higher is better.
Monthly pricing
Standard USD rates, billed month to month. Annual discounts and promotions excluded.
-
Monthly — 20,000 words
$19.00/mo20,000 humanizer words per month.
- Larger word allowances are available below.
-
Unlimited — web
$290.00/moUnlimited humanizer words advertised.
- API access is not included.
-
Unlimited — with API
$990.00/moUnlimited words with API access advertised.
Free access and trials
A basic humanizer model is free; its recurring allowance is not published. A separate premium trial includes 250 words.
- Word allowances are configurable. Unlimited detection is a separate feature and does not make metered humanization unlimited.
- Input/output accounting and plan-specific request caps are not published on the pricing page.
Sample output
One AI-written passage from the October 2026 cycle and what Undetectable.ai returned for it. This prompt was picked by a fixed rule, not by how the tool scored.
AI-written input
Every tool on this prompt →Discussion Post · Discussion Board · written by claude-sonnet-5-5 · 274 words
Communities should treat public art and infrastructure as complementary investments rather than competitors, while still giving priority to safety and basic function when budgets are tight.
Start with the principle that essential infrastructure comes first. Water systems, roads, bridges, and emergency services protect lives and underpin economic activity. When a community faces deferred maintenance or safety hazards, diverting money to art is hard to justify. Residents rightly expect core services before embellishments.
That said, art is not a frivolous add-on. Research and local experience show that public art can strengthen identity, attract visitors, support local artists, and make neighborhoods feel cared for. The most effective approach is integration. Many cities adopt "percent for art" policies, setting aside one to two percent of capital project budgets for artistic elements. This keeps art spending proportional, ties it to projects already underway, and avoids pitting it directly against repair needs. A new bridge, transit station, or plaza can include thoughtful design at modest marginal cost.
Transparency and public participation also matter. Communities should involve residents in deciding what gets funded, using open budget hearings and clear criteria. People are more supportive of art when it reflects local history and values, and when they see infrastructure needs being met first.
Finally, diversify funding. Grants, private donations, business sponsorships, and arts endowments can finance projects without drawing on road or utility budgets. Temporary installations and community murals offer low-cost options.
In short, fund essentials reliably, embed art in capital projects through small set-asides, seek outside money, and let residents help set priorities. This balance respects both practical needs and the cultural life that makes a place worth living in.
Show the full input Show less
Undetectable.ai output
279 words (input 274)
While safety must be the top priority, investments in public art and infrastructure should be viewed as complementary in community building.
First, ensure the basic infrastructure required to sustain life and support economic development is in place and maintained (water systems, roads, bridges, emergency services etc.). Only after addressing any issues of deferred maintenance that threaten safety should consideration be given to how to support public art using available funds.
Incorporate public art in important projects by designating one to two percent of the funds for a public art component. People are surprised at the modest amount of additional funds that thoughtfully designed public art can cost as an add-on to, say, a new bridge, a major transit station, or a major plaza redevelopment. In all of these cases the artistic design should be an integral part of the capital budget for the project.
If you commit your residents to be part of deciding how their dollars will be spent, using open budget hearings and clear criteria, then they will be more supportive of the resulting public art if it contains references to local history and their values. Then you can make sure that basic infrastructure needs are met as well.
Use multiple funding sources. Use grants, donations, business sponsorships, and a community arts endowment fund. Utilize temporary works of art and utilize community murals for low-cost options.
Funding essentials, including meaningful sets-asides in capital projects to include art in public buildings and spaces, the pursuit of outside funding (grants, private donations, business sponsorships, etc.), and neighborhood input regarding priorities for public art will sustain both the core of a community and the full richness of cultural life there.
Show the full output Show less
- GPTZero1.00passed
- Pangram0.00caught
- Winston AI1.00passed
- ZeroGPT0.74passed
- Copyleaks1.00passed
- Originality.ai1.00passed
Passed 5 of 6 detectors.
Every Undetectable.ai run in October 2026 →Each chip is one detector's score for that output, from 0 (AI) to 1 (human). A score of 0.50 or higher counts as a pass and shows green; below that the detector caught it and the chip shows red.
Penalties applied this cycle
How penalties work →The overall score shown above already reflects these deductions.
- Length inflation ×5−5.0
The output ran much longer than the input, a common trick that pads text to dilute the AI signal.
Compare with
Featured in
Questions about Undetectable.ai
Answers are generated from the benchmark data and update every month. Last updated with the October 2026 cycle.
Is Undetectable.ai good?
Undetectable.ai ranked #3 of 14 in the October 2026 HumanizerBench cycle with an overall score of 71.18/100, above the field median of 66.78. On the parts of the score, it placed #5 of 14 on detector bypass, #2 of 14 on meaning preservation and #5 of 14 on readability. The score includes a 5-point deduction for length inflation.
Does Undetectable.ai bypass AI detectors?
In the October 2026 cycle, Undetectable.ai had a detector bypass rate of 74.6% (#5 of 14 tools). 1 of its 30 outputs passed all 6 detectors, and 29 were caught by at least one. Its average human-likelihood score from each detector, where 100 means every output was scored as human: Winston AI 98.5, Originality.ai 88.7, Copyleaks 87.9, GPTZero 85.0, ZeroGPT 83.8 and Pangram 3.6.
Can Undetectable.ai bypass GPTZero?
In the October 2026 cycle, Undetectable.ai's average human-likelihood score from GPTZero was 85.0 out of 100 (#6 of 14 tools against GPTZero). The highest average GPTZero score that cycle was 99.5, by WriteHuman. See every tool's GPTZero results.
Is Undetectable.ai free?
A basic humanizer model is free; its recurring allowance is not published. A separate premium trial includes 250 words. Paid plans start at $19.00 a month, billed monthly.
Is Undetectable.ai legit?
Undetectable.ai is a working AI humanizer at undetectable.ai. HumanizerBench has tested it in 5 monthly cycles, most recently October 2026, when it completed 30 of 30 test runs. Every input and output from its runs is published with each detector's verdict.
Does Undetectable.ai keep the original meaning?
Undetectable.ai scored 79.2 out of 100 for meaning preservation in the October 2026 cycle (#2 of 14; field median 71.0). The score measures how close each rewrite stays to the meaning of the original text. Its readability score was 63.9 (#5 of 14). It lost 5 points to penalties for length inflation.
Which kinds of writing does Undetectable.ai handle best?
In the October 2026 cycle, Undetectable.ai did best on application essays (76.4/100) and worst on blog posts (58.1/100). On academic essays it ranked #6 of 14.
How has Undetectable.ai ranked over time?
Undetectable.ai has been ranked in 5 HumanizerBench cycles: #5 in June 2026, #2 in July 2026, #8 in August 2026, #6 in September 2026 and #3 in October 2026. Its overall score went from 80.09 in June 2026 to 71.18 in October 2026.
What are the best alternatives to Undetectable.ai?
In the October 2026 cycle, the tools ranked above Undetectable.ai were WriteHuman (#1, 74.89/100) and StealthGPT (#2, 74.50/100). Compare Undetectable.ai with any of them side by side.
Badges
Undetectable.ai is ranked in the October 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/undetectable/october-2026.svg, which renders October 2026 forever.