November 2026 · Independent humanizer benchmarks. Explore the evidence

Terms of this preview

Research results, with limits

The data describes the saved outputs and recorded measurements in this run. Scores are not a guarantee of future product performance, human authorship, academic acceptance, or detector behavior on unseen text.

Data attribution

The published source texts and historical service rewrites are attributed to AI Humanizer Benchmark under CC BY 4.0. The original scoring code is MIT-licensed. Model artifacts retain their own upstream licenses.

Local preview

This preview runs on localhost. The planned humanizebench.com address is branding and canonical metadata; this build does not claim that domain has been deployed or registered.