# Rhythm uniformity and burstiness > Source-anchored research for Rhythm uniformity and burstiness. Run status: COMPLETED. Parent: [LLM rhetorical-pattern tells in prose](https://llms-explorer.com/tree/ai-isms-rhetorical-tells/) · 2 facets · 5 facts · page: https://llms-explorer.com/tree/rhythm-uniformity-and-burstiness/ ## Definitions - Burstiness is defined as how much writing patterns and perplexities vary over a whole document; the claim is that humans vary naturally while LLMs choose next words by the same rule and so show low burstiness. (medium confidence) — [source](https://gptzero.me/news/perplexity-and-burstiness-what-is-it/) ## Facts and statements - Across six LLMs from three families, human news text shows more scattered sentence-length distributions, more varied vocabulary and dependency types than LLM text; differences between humans and LLMs exceed differences between LLMs. (low confidence) — [source](https://arxiv.org/abs/2308.09067) - GPTZero stopped relying on perplexity and burstiness as primary detection signals in autumn 2023, moving to a deep-learning classifier; the stats remain one of seven indicators. (medium confidence) (verified as of 2026-09-23) — [source](https://gptzero.me/news/perplexity-and-burstiness-what-is-it/) - Perplexity/burstiness are relative to a specific scoring model, so results shift with each model release; closed models expose no token probabilities, forcing open-model proxies. (low confidence) — [source](https://www.pangram.com/blog/why-perplexity-and-burstiness-fail-to-detect-ai) - Text heavily present in training data (e.g. the Declaration of Independence) scores as low perplexity and is misflagged as AI; non-native English writing also scores low perplexity/burstiness (2023 Stanford TOEFL study), so the metric produces false positives. (low confidence) — [source](https://www.pangram.com/blog/why-perplexity-and-burstiness-fail-to-detect-ai)