AI Tells

"AI tells" are the recurring habits that make machine-written text recognisable — to attentive readers and, in aggregate, to detectors. They aren't proof on their own, but stacked together they form a signature that's hard to miss once you know what to look for.

The vocabulary tells are the most famous: an over-fondness for words like "delve," "tapestry," "realm," "testament," and "underscore," and connective phrases like "it's important to note," "in today's fast-paced world," and "navigating the landscape of." Models reach for these because they're high-probability continuations, which is exactly why they read as generic.

The structural tells are subtler and matter more for detection: uniform sentence length, a fondness for tidy three-item lists, symmetric paragraphs, an even and slightly formal register, and frequent em-dashes used in a consistent way. Human writing is messier — it varies its rhythm, drops articles, and takes tangents. That messiness is the burstiness detectors look for.

Removing AI tells is partly editing and partly humanizing. You can hand-edit the obvious vocabulary, but the structural uniformity is what a humanizer addresses systematically — SynthGuard's phrase-replace layer swaps the predictable phrasings and the sentence-restructure layer breaks the flat cadence. Fixing tells improves readability, but it doesn't guarantee any specific detector outcome, and the strongest signal of human authorship is still genuine, specific content only you can add.