AI Text Pattern Detector

What this actually measures — and what it cannot

This tool looks for surface patterns associated with machine-generated prose: stock transitional phrasing, formulaic hedging, a particular register of evaluative adjective, and structural tics like tidy triads and summary sentences that restate the paragraph just given.

It is important to be clear about the limit. This is not a detector and cannot be one. There is no reliable way to determine whether text was machine-written from the text alone — published detectors have high false-positive rates, and they are documented to misfire disproportionately on writing by non-native English speakers, whose more formulaic constructions look like the patterns being flagged.

Treat the output as a style observation about the text in front of you, never as evidence about its author. Using a score like this to accuse someone is not a use this tool supports, because the underlying inference does not hold.

The legitimate use

What it is genuinely good for is editing your own drafts. The patterns it flags are worth removing regardless of who wrote them: "it is important to note that", "in today's fast-paced world", paragraphs that open by announcing what they will say and close by repeating it. That prose is flabby whether a person or a model produced it.

Used that way the score is a proxy for something real — how much of the text is scaffolding rather than content — and cutting the flagged constructions almost always tightens the writing.