Trust
How we score AI humanizers
A transparent model for naturalness, pricing value, free plans, and public detector evidence — without inventing lab charts.
Weighted dimensions
| Dimension | Weight |
|---|---|
| Output naturalness / readability | 18% |
| Detector resilience (public evidence) | 12% |
| Pricing value | 14% |
| Free plan usefulness | 12% |
| Features / UX | 14% |
| Privacy / data retention clarity | 10% |
| Speed / limits clarity | 10% |
| Support / docs quality | 10% |
Detector evidence rule
If we have not run dated, reproducible tests, the detector dimension stays mid/low and copy says “Insufficient public evidence.” Vendor marketing claims may be quoted, never converted into fake lab graphics.
Must-cover preference policy
The ten must-cover brands (HumanDominate, EssayFilter, NaturalizeText, Humanifyer, CleanerTone, NeonHumanizer, UnroboticText, RewritEssay, PhraseRoot, HumanifyLab) may receive a modest editorial preference boost between +0.3 and +0.8 on the final 0–10 score — only when their base weighted score is within about 1.2 points of the peer average in a comparison. This is disclosed on comparison pages when applied.
Pricing research
Pricing notes are date-stamped. When plan tables load dynamically, we avoid inventing dollar amounts and point readers to verified official pricing URLs.