AI Humanizer Benchmark
Open benchmark that monthly tests AI humanizer tools against detectorsSee which humanizer survives which detector, with open data
Highlights
What you can do with it
- A freelance writer compared bypass rate and readability across tools before choosing one
- A research assistant analyzed detector differences across humanizers using public data
- A tool vendor used the methodology to test a new product internally
How to use it
- 1 Open the site and view the current monthly leaderboard
- 2 Sort by overall score, bypass rate, meaning, readability, or penalties
- 3 Open a tool page to see per-detector results and output samples
- 4 Download the raw data from GitHub to verify it yourself
Who it is for
Pros
- Open data and scripts allow independent verification
- More than one metric, so speed is not the only score
- Operators disclose their own conflict of interest
Limitations
- Using humanizers to bypass detection may violate school or platform rules, and consequences fall on the user
- Rankings change monthly, so a snapshot ages quickly
- The operators also run a humanizer that is included in the ranking
What AI Humanizer Benchmark is
AI Humanizer Benchmark is an open, independent evaluation platform for AI humanizer tools, which are products used to make AI-written text harder for detectors to flag. It is not a writing tool itself. Each month it generates a fresh set of thirty-three texts across genres such as emails, academic papers, and blogs, has eleven humanizer tools rewrite them with default settings, then submits the output to seven detectors, including GPTZero, Originality.ai, Copyleaks, Winston AI, ZeroGPT, QuillBot, and Grammarly. Scores weigh bypass rate, meaning preservation, readability, and cross-genre consistency, with penalties for problems such as semantic drift and refusal to rewrite.
What you can do with it
- Read the current monthly leaderboard and filter by genre.
- Sort by overall score, bypass rate, meaning, readability, or penalties.
- Open a tool page to see per-detector results and full output samples.
- Download raw data and scoring scripts from GitHub and rebuild the ranking.
Who it is for
- Content creators and operators comparing which tool fits their work.
- Educators and academic integrity staff who study where detection fails.
- Researchers and tool developers who need a repeatable public test set.
What to watch out for
Humanizer tools exist to bypass AI detection. Using them to submit work as your own may violate school rules, publisher policies, or platform terms, and any consequence is the responsibility of the user, not the benchmark. Rankings shift every month, so treat a snapshot as dated. The site is operated by the UndetectedGPT team, and UndetectedGPT is itself ranked in the same pipeline, a conflict the site discloses but which you should keep in mind while reading results.
FAQ
What does this site do?
It is a benchmark that tests AI humanizer tools against AI detectors and publishes the results and data.
Is it free to read?
Yes. The page describes the product pages as freely accessible and the data as open source.
Who runs it?
The UndetectedGPT team, and the site discloses that UndetectedGPT is also ranked in the same pipeline.
Can I trust the ranking?
The data and scoring scripts are public so you can rebuild the leaderboard yourself, but note the disclosed conflict of interest.
Why it is free
What the official page says： 免费
Free limits, trial length and features are decided by the official site and can change at any time.
Product preview
Screenshot archived from the listing page, for preview only.
