Humanization Benchmark
Same argument.
Three voices.
One essay on AI and humanity, written once by a model, then rewritten by two humanization skills. Scored blind, category by category, by OpenAI 5.0.
Synthetic → Human spectrum
Positioned by overall score
0 · Synthetic
10 · Human
Results
Leaderboard
Ranked by overall score across all three rubric categories. Tap a card to read that sample in full.
Shape of the score
Category radar
The same three categories, plotted per sample — the fuller the shape, the stronger the sample.
The rubric
Category breakdown
What each category actually measures, and how every sample scored against it.
Read them yourself
The full samples
Every word each sample was actually scored on — judged by OpenAI 5.0.