Skip to content
Glossary

Zero-Shot Detection

A plain-language explanation for writers, not engineers.

What It Means

An AI detection approach that works without being trained on specific examples of AI-generated text. Instead of learning patterns from a labeled dataset, zero-shot detectors use the statistical properties of language models themselves to estimate the probability that text was AI-generated.

Why writers should care: Zero-shot detectors are the most common type - and the most prone to error, because they're making educated guesses rather than matching known patterns.

In Context

Zero-shot detection is essentially asking one AI model to guess whether another AI model produced a given text. No labeled training data, no examples of known AI text - just statistical inference. It is the most common approach because it requires no dataset curation, but it is also the least reliable. Studies show zero-shot methods achieve 60-80% accuracy at best, with false positive rates that make them unsuitable for high-stakes decisions. Every time a zero-shot detector is used to decide a student's grade, an unreliable guess is treated as evidence.

Related Terms

← Back to the Full Glossary

The Sunday Letter

Every Sunday, one email. A featured essay, a case study update, a craft tip, and a writing prompt. No AI wrote this.