The Mechanics of AI Detectors Explained
A deep dive into how Turnitin, GPTZero, and other detectors analyze perplexity and burstiness.
SJ
Sarah Jenkins
October 18, 2026 • 8 min read
Understanding how AI detectors work is the first step to ensuring your authentic writing doesn't get falsely flagged. Most modern detectors don't just look for specific words; they analyze the statistical properties of the text.
Perplexity and Burstiness
The two main metrics used are perplexity and burstiness. Perplexity measures how predictable your text is to an AI model. If the AI easily guesses the next word, perplexity is low (more likely AI-generated). Burstiness measures the variation in sentence length and structure. Humans write with high burstiness; AI tends to be monotonous.
