Perplexity and Burstiness Explained for Normal People

AI detector language can sound more scientific than it feels in practice. Perplexity and burstiness are two common ideas behind detection, and both are really about the same question: does this text move like a human wrote it?

Perplexity Means Predictability

Low-perplexity text is easy for a language model to predict. It uses expected words in expected places. AI writing often has low perplexity because the model is literally trained to choose likely next words. Human writing tends to include more surprising choices: an odd example, a sharper verb, or a less obvious sentence turn.

Burstiness Means Variation

Burstiness describes how much sentence length and complexity change across a piece. Humans write in bursts. A short sentence lands. Then a longer one explains the reason, adds a caveat, or pulls in context. AI drafts often keep every sentence in the same comfortable range.

Why This Matters for Editing

Do not chase detector math directly. Edit for clarity and human usefulness: stronger nouns, less filler, better examples, and a more natural rhythm. Those changes often improve the same signals detectors are trying to measure.