AI-Tell Detector

You know the feeling. You read a paragraph, it is grammatically spotless, nobody is technically wrong, and yet something in you goes: a machine wrote this. This little tool is that feeling, turned into a checklist. Paste any prose and it flags the vocabulary, the constructions and the rhythms that read as machine-made, built from the strip-the-AI-tells rules I use on my own book, plus the current crop of AI-darling words (delve, tapestry, testament, the whole choir). Drama is allowed. The tell is the tidy symmetrical machinery underneath it. No model, no API, nothing leaves this page: it is just you, your paragraph, and some judgmental highlighting.

How to read this. This is a tells highlighter, not a lie detector, and definitely not a court of law. A high score means the writing carries patterns common in AI drafts. It does not prove a machine wrote it, and confident human writing can trip a flag or two (mine does, regularly, which is humbling). Use it to spot and revise, never to accuse. And if your score comes back high, do not panic. Nobody is coming for you. Fix the three worst offenders, run it again, and watch the number drop. Every rule is plain text in the code below, yours to edit.

Enjoyed this? It comes from the thinking behind my book, Human Still Wins