What Is Burstiness?
Definition
The variation in sentence length and complexity throughout a piece of writing. Human text is naturally bursty — mixing short, punchy sentences with long, complex ones. AI text tends to be uniform, with sentences of similar length and structure.
Burstiness Explained
Burstiness measures the degree to which sentence complexity varies within a document. The concept originates from information theory and was adapted for AI detection by researchers who observed that large language models produce remarkably uniform text. When a human writes, they naturally alternate between simple declarative statements and elaborate constructions with multiple clauses, parenthetical asides, and varying rhythms. This creates a bursty pattern — periods of simplicity punctuated by complexity. AI models, trained to maximize probability at every token, tend to settle into a comfortable middle ground of moderate complexity that rarely deviates. A 2,000-word essay written by a human might contain sentences ranging from 4 to 45 words, while the same length AI essay typically stays within a 12-to-25-word range. Burstiness is measured alongside perplexity and serves as a complementary signal — even if perplexity appears normal, low burstiness alone can trigger detection.
How Burstiness Relates to AI Detection
GPTZero explicitly measures burstiness as one of its two core signals (alongside perplexity). Turnitin analyzes sentence-level variation as part of its classification features. Low burstiness is especially telling for longer documents where the uniform pattern becomes statistically significant — a 200-word paragraph might not show it, but a 2,000-word essay almost certainly will.
How Anti-Turnitin Handles Burstiness
Anti-Turnitin's post-processing engine deliberately introduces natural burstiness patterns. After the LLM rewrite, the algorithm varies sentence length and complexity to match human writing distributions — inserting occasional short fragments, combining clauses into longer constructions, and varying paragraph density to create the bursty rhythm detectors expect.
Related Terms
Perplexity
A measure of how predictable the next word in a sequence is. Low perplexity means the text follows expected patterns closely, which is a hallmark of AI-generated writing. Human writing typically has higher, more variable perplexity.
Entropy (in AI Detection)
Shannon entropy measures the randomness or information content in text. In AI detection, low entropy indicates predictable, formulaic writing typical of language models, while higher entropy suggests the varied word choices characteristic of human authorship.
Text Classifier
A machine learning model that categorizes text into predefined groups. In AI detection, text classifiers are trained to distinguish between human-written and AI-generated text by analyzing statistical features like perplexity, burstiness, and token distributions.
Frequently Asked Questions
How is burstiness different from perplexity?
Perplexity measures word-level predictability — how surprising each word is in context. Burstiness measures sentence-level variation — how much the complexity changes from one sentence to the next. They are complementary signals; AI text tends to be low in both.
Can I add burstiness by just varying sentence length?
Artificially varying sentence length helps, but detectors measure complexity beyond just word count. True burstiness includes variation in syntactic depth, clause nesting, vocabulary sophistication, and information density. Simply writing one short sentence followed by one long one in a mechanical pattern can itself appear artificial.
See how Anti-Turnitin handles these signals
Try Anti-Turnitin FreeNo credit card required. Paste your text and get results in under 3 seconds.