Glossary
Plain-language definitions of the terms writers run into.
- Perplexity
- Perplexity is a measurement of how surprised a language model is when it encounters a piece of text — technically, how unpredictable each next word turns out to be given the words before it. Text that follows the statistical grooves of typical writing (common phrasings, expected word orders) has low perplexity; unusual, creative, or irregular writing has high perplexity.
- Burstiness
- Burstiness describes the degree of variation in sentence length and structure within a piece of writing. High burstiness means the text alternates between short and long sentences unpredictably — a twenty-word sentence followed by a four-word one. Low burstiness means sentences cluster around similar lengths with regular rhythm. Human writing, measured across large samples, tends toward higher burstiness; machine-generated text clusters at uniform lengths because models resolve each next-sentence decision toward the statistically average choice.
- Patchwriting
- Patchwriting is the practice of restating a source text by replacing some of its words with synonyms while keeping the original sentence structure largely intact. The result sits in an uncanny middle zone: not a direct quote, not a genuine restatement. Howard's classic description calls it 'copying from a source text and then deleting some words, altering grammatical structures, or plugging in one-for-one synonym-substitutes.'
Why these terms matter
Detector vocabulary is worth learning because flags arrive without explanations: a report says “likely AI-generated” and leaves you to work out what triggered it. Knowing what perplexity and burstiness actually measure lets you read that report critically — to see why formulaic human prose gets flagged, and what evidence would actually change the verdict. The same vocabulary doubles as an editing language: once you can name uniform sentence rhythm or a synonym-swapped paraphrase, you can fix those traits deliberately instead of shuffling words at random. The AI humanizer applies these concepts directly, restructuring rhythm and phrasing rather than masking text — the difference between reacting to a score and understanding your own prose.