Burstiness
The variance in sentence length and structure — human writing is bursty, AI writing is uniform.
Burstiness measures the variance in sentence length and structure across a passage. Humans naturally write with high burstiness — short punchy sentences mixed with long flowing ones, fragments next to compound-complex constructions. LLMs default to a middle-ground sentence length and stick close to it, producing low burstiness.
GPTZero scores burstiness directly and weighs it heavily. Originality.ai uses it as one of several stylometric features. Most academic AI detectors use a burstiness derivative.
SynthGuard's sentence-restructure layer raises burstiness by splitting long sentences at clause boundaries and merging short ones via conjunction. The target is a standard deviation of sentence length around 10–14 words, which puts a passage safely in the human band.
The tool ships a live burstiness meter so you can watch your score climb as the layer applies. Once you're above ≈0.55 on GPTZero's scale, the AI-classification probability drops sharply.