AI-tell score
Breakdown
What it checks
- Buzzwords
- Certain words show up in machine writing at a rate way past normal speech. One on its own means nothing. A cluster of them in one paragraph is the tell.
- Filler phrases
- Stock openers and closers a model falls back on: a throat-clearing lead-in, a chipper offer to help at the end, an announcement before diving into the topic. They pad the word count without adding anything.
- Constructions
- Sentence shapes, not single words: setting up a contrast between two things just to land on the second one, restating a fact by first denying its opposite, and opening with a question purely as a hook instead of getting to the point.
- Hedges
- Words like can, often, and typically, stacked up in one stretch of text. Not scored on its own, but worth a look. Too many hedges in a row reads evasive.
- Em dashes
- An occasional dash is normal punctuation. A pileup of them, one in nearly every sentence, is a habit worth breaking.
- Emoji in prose
- Fine in a text message, out of place in a report, a README, or a cover letter.
- Bold-label bullets
- A long run of bullets that all follow the same shape: a bolded word, a colon, then a short explanation. One or two is a normal list. Four or five in a row is a template.
- Sentence rhythm
- Real writing varies its sentence length without trying to. When every sentence lands within a word or two of the same length, that evenness is its own tell.