Check your text. Then fix it. Then check again.

How to read the score like an adult

The detector returns a score from 0 to 100 representing AI likelihood, and unlike most detectors, it admits the middle exists. Below 40, your text reads human: the statistical patterns sit comfortably inside normal human variance. Above 70, it likely reads AI generated: the uniformity and predictability tip past what people produce naturally. Between 40 and 69, the honest answer is inconclusive, and that is exactly what the tool says. Plenty of human writing lands in that band, especially formal prose, and plenty of lightly edited AI text does too.

Every score on this site ships with the same disclaimer: a detector reading is a signal, not a verdict. That is not legal boilerplate; it is how the technology actually behaves. The same text scores differently across detectors. Short texts produce unstable scores, which is why we require 120 characters minimum. Non-native English writers get falsely flagged at documented, higher rates. Anyone who treats a single detector number as proof of misconduct has misunderstood the tool, and we say so on the page where the number appears.

The loop nobody else gives you on one site

Detection alone tells you there is a problem. This site also hands you the fix. Score your draft, and if it comes back high, click through to the humanizer: your text rides along automatically. Humanize, return, re-check. Watching your own score drop from 85 to 30 is more convincing than any marketing claim we could write, and it costs you nothing.

For the mechanics behind the number, perplexity, burstiness and classifier training, read how AI detection works. For how the big institutional detectors behave, see the Turnitin guide and the GPTZero evidence.

When to bother checking at all

Not every text needs a detector. Check when a third party will judge the text's origin: submissions, applications, published work under your name. Skip it for private notes and low stakes messages, where the only reader who matters is the recipient. And never use any detector, ours included, to accuse someone else; the false positive evidence makes single score accusations indefensible, and the people most likely to be falsely flagged are the ones writing carefully in their second language.