Why no AI detector can be trusted as a verdict
This looks for three surface patterns sometimes associated with AI-generated text: unusually uniform sentence lengths, a handful of stock transition phrases that show up disproportionately in AI writing, and the absence of contractions. Every one of these also occurs naturally in plenty of human writing - a careful human writer, a non-native English speaker following rules learned in class, or simply a particular personal style can all trigger the same signals.
The three signals, explained
"Burstiness" measures how much sentence length varies across a passage - human writing tends to mix short punchy sentences with longer, more complex ones more than AI-generated text does on average, so unusually uniform lengths raise this signal. Stock phrases ("it is important to note," "in today's fast-paced world," "delve into") are expressions that show up disproportionately often in AI-generated text, based on widely discussed public observations, though plenty of human writers use some of these phrases too. The absence of contractions is the weakest of the three signals on its own, since many formal writing styles deliberately avoid contractions regardless of who - or what - wrote the text.
What independent research actually shows about AI detectors
Independent research on commercial AI-detection tools has repeatedly found meaningful false-positive rates, including against real human-written text - and these false positives disproportionately affect non-native English speakers, whose writing style can trigger the same pattern-matching signals these detectors rely on.
What this tool should never be used for
Treat this as a curiosity about your own or someone else's writing style, never as evidence. It should never be used as grounds to accuse a student of academic dishonesty or an employee of professional dishonesty - doing so risks serious, unfair harm based on a score that measures surface pattern matches, not a reliable finding about who or what actually wrote a piece of text.