Paste a baseline piece (something on-voice) + a candidate piece (the ghostwriter / AI-assisted draft you're checking). We compare seven stylistic axes — formality, sentence length, sentence rhythm, vocabulary richness, first-person pronouns, hedging, jargon density — and return a 0-100 consistency score with per-axis drift detail. Pure client-side, no LLM, nothing leaves your browser.
Formal writing uses longer, Latinate words. Casual writing prefers short, Anglo-Saxon ones. We measure average word length in both pieces and compare — a big jump signals the candidate is either much more corporate or much more casual than the baseline.
Voice isn't just the words — it's the cadence. A punchy writer varies sentence length deliberately: short. Then longer and more expansive. Then short again. Average sentence length tells you speed; variance tells you whether that rhythm pattern is preserved.
The type-token ratio — unique words divided by total words — measures how diverse the vocabulary is. A writer who repeats the same words will score low; one who rarely reuses vocabulary scores high. Large divergence between pieces means one is much more or less lexically varied.
These three axes capture personality signals that are hard to copy correctly. Some writers say "I" constantly; others avoid it. Some hedge everything ("probably", "maybe", "it seems"); others assert. Industry jargon at the wrong density is the most common tell on ghostwritten content.
A score of 80–100 means the candidate is stylistically indistinguishable from the baseline across the seven axes we measure. Below 60, at least two axes are significantly off — enough that a reader who knows the author's usual cadence would notice something is different. Below 40 indicates the pieces read as if written by different people.