How we compile entries

Our process for documenting AI humanizers and detectors.

Humanizer Wiki is a reference, not a review site. We don't score or rank tools. We record what each one does, how it's priced, and how it fits the wider landscape, then link it all together.

What each entry records

• What the tool is: its category, purpose, and how it works.
• Pricing: plan structure and whether a free tier exists, as stated by the vendor.
• Capabilities: features, and what it does and doesn't do well, in descriptive terms.
• Relationships: which detectors a humanizer is measured against, and related tools and research.

How we test

1. We run a fixed set of AI-generated passages (short, medium, and long-form) through each tool.
2. We re-check outputs with multiple detectors, including GPTZero and Originality.ai.
3. We read the output ourselves and describe what we observe, without reducing it to a score.
4. We date every entry, because tools and detectors change often.

Sources

Entries draw on vendor documentation, hands-on testing, and published research. Where a claim comes from a vendor (for example, an accuracy figure), we attribute it as such rather than presenting it as our own finding.

Independence & disclosure

Nothing here is for sale. Some outbound links may be affiliate links that earn us a commission; this never affects what we document, and affiliate status is disclosed. Entries still being verified carry a notice at the top.

An honest limitation

Detection is probabilistic and adversarial, so any observation is a snapshot in time. We revisit entries and update the "last updated" date. See Can humanizers actually beat detectors? for the fuller picture.