Terms of use

Anyone may read the site, link to any page, and share any page address. There is nothing to sign up for and nothing to agree to before reading. The site is for people deciding whether to use an AI humanizer and which one; it does not give legal advice, academic advice or advice on any institution's rules, and a reader who needs those should ask the institution.

Automated access that reads pages at a rate a person could not, or that copies the site wholesale, is not permitted. Reading the published data files by script is permitted; they are published for that.

Any sentence, table row or verdict on the site may be quoted, in print or online, with two conditions: name Best AI Humanizer as the source, and link the page the quote came from. A quote should carry the date of the page it came from, because every price and reading on the site is true for its date and may have changed since. A quote may not be edited to change its meaning, and a verdict may not be presented as applying to a tool or a plan it does not name.

Vendors may quote their own review on the same terms, including the main limitation, and may not quote part of a sentence in a way that reverses it.

Every file a run publishes for download, including inputs.csv, outputs.csv, detector-readings.csv, ratings.csv and the screenshot archive, is released under the Creative Commons Attribution 4.0 International licence. You may copy, share, analyse and build on the files, including for commercial work, as long as you credit Best AI Humanizer, link the run page the files came from, state the run ID, and say whether you changed anything. The licence deed at the link above is the full text; nothing on this page narrows it.

The licence covers the site's own work: the human control texts, the ratings, the readings and the screenshots. The AI-written inputs are published as test material; the output texts in outputs.csv were produced by the tools under test from those inputs, and a reader reusing them should bear that in mind. The method that produced the files is on how we test AI humanizers.

A detector reading is the reading that detector gave on the stated date; detector models change without notice. A meaning rating is the judgment of two raters under a published rubric, with both scores published. A score is the weighted sum of those, under weights published before the run. None of them is a guarantee that any tool will produce the same result on another text, on another day or on another detector, and no page on the site states that a tool makes text undetectable or passes any named detector.

The site is offered as is. The editors check every fact against its source and log every correction in the corrections log, and they make no warranty that the site is free of error or that any fact remains true after the date on it. The site is not responsible for a decision taken on the basis of a reading, and no detector score should be the only evidence in a misconduct case.

Questions about these terms, requests to reuse material in a way they do not cover, and reports of a quote used against them go to the editors through the contact page. The privacy policy covers what the site collects, the about page describes the editors, and the bestaihumanizer.ai list of twelve tools is where the reviews and the data begin.