AI Text Pattern Checker
This does not detect who or what wrote your text, because nothing reliably can. It measures the patterns that make prose read as generated, so you can edit them out.
Note: Never use a result from this — or from any detector — to accuse someone of anything. Published detectors misclassify human writing often enough that several universities have withdrawn them, and they misfire hardest on people writing in a second language.
What it does instead is measure the specific patterns that make prose read as generated, whoever wrote it. Openers that spend a sentence clearing their throat before the point. Sentences of near-identical length, paragraph after paragraph, which is what produces the flat rhythm people describe as robotic. Stock transitions — moreover, furthermore, in conclusion — clustered more densely than any editor would leave them. Hedging that removes all commitment from a claim. Lists of exactly three adjectives. Every one of those is a fixable editing problem, and each is reported with the sentences it found so you can go and change them.
Sentence length variance is the measure worth watching. Good prose varies: a long sentence carrying a qualified argument, then a short one landing it. When the standard deviation drops below about four words, the writing has that even, unaccented quality no reader consciously notices and every reader finds tiring.
Use it as an editing pass, not as evidence. A high score means the text needs work. It does not mean a machine wrote it, and it should never be used to accuse anyone of anything.
Frequently Asked Questions
Can this prove text was written by AI?
No, and nothing can. Detection tools produce false positives at rates that make them unsafe for any consequential decision, and they are disproportionately wrong about people writing in a second language. Treat the output as an editing checklist, never as evidence about who wrote something.
What sentence length variance should I aim for?
A standard deviation above about six words reads as natural, and below four reads as flat. The fix is not to add long sentences but to add short ones — a five-word sentence after a thirty-word one creates the rhythm change readers respond to.
Why are common transition words flagged?
Not because they are wrong, but because their density is a reliable signal. One 'however' in a page is invisible; four in three paragraphs is the pattern that makes writing feel assembled rather than thought through. The count matters more than the word.
Does my text get uploaded or stored?
No. The analysis is plain string and statistics work running in your browser, with no request to any server. That is worth knowing given the kind of draft people paste into a tool like this.
