Independent · Updated monthly
Humanizer Audit
AI humanizers promise text that slips past detectors. We read every output, blind, to see which ones still say what you wrote.
blind-review.diff
Bottom line
Ranking
All benchmarks
Across the benchmarks
Combined
One ranking from all three
What matters to you
How much each benchmark counts
Combined ranking with these settings
Who holds up in the combined ranking
Detection versus faithfulness
Where each tool lands
Tool by tool
Rank it your way
Ranking with these weights
Who holds up across weightings
Over time
Six months of the same pattern
Usability in the blind review, 1 to 5, for every audited month. A gap means the tool was not in that month's test.
The leaderboard's #1 against our faithfulness ranking
Every audited month
Detector results
Its own quality checks
Share of each tool's outputs that passed the check, as HumanizerEval reports it. Factual consistency and naturalness reproduce from its per-output file; the other checks have no per-output data and are its own claim.
Other benchmarks
Is there anything better?
Method and limits