AI Detector
A tool that estimates whether a passage of text was generated by an AI model, by measuring statistical properties of the writing rather than by checking any record of its origin.
Why it matters
AI detectors are used to decide whether students pass, whether freelancers get paid and whether writers keep clients — which is far more weight than the published research supports. Understanding that they measure how text reads, not who wrote it, is the difference between using one sensibly and using one to make a decision you cannot defend.
A concrete example
A student submits an essay written entirely by hand. Because English is their second language, their vocabulary is simpler and their sentence construction more regular than a native speaker's — the same properties detectors read as machine-like. The tool returns 92% AI-generated, and the student now has to prove a negative about their own work.
How to use it
Treat the output as a signal that starts a question, never as a finding that ends one. On your own drafts it is a reasonable editing prompt: a high machine score usually means the text is still unedited and flat, which is worth fixing for reasons that have nothing to do with detection. Across a batch of commissioned content, a high flag rate is a reason to look more closely at a supplier. What it cannot support is a decision about an individual, because the false positives are frequent and unevenly distributed.
The common mistake
Reading the probability as a measurement of origin. A detector has no access to where text came from — it compares statistical properties against what it expects from human and machine writing and reports a likelihood. Detectors misclassify non-native English writing as machine-generated at high rates, which means the people most likely to be wrongly accused are often the least able to argue back.
Related terms
Perplexity
A measure of how surprising a piece of text is to a language model — roughly, how unexpected each word is given the ones before it. Low perplexity means the text follows predictable paths.
Burstiness
The variation in complexity and length across the sentences of a passage. Human writing tends to be bursty — long sentences next to short ones — while generated text tends to be more uniform.
Plagiarism Detection
Automated systems that compare text against existing published content to identify similarity. Critical for AI-generated content, which can inadvertently reproduce training data.
AI Alignment
The research challenge of ensuring AI systems pursue goals that are beneficial to humans. Misaligned AI could technically achieve its objective while causing unintended harm. Alignment research aims to make AI reliably helpful, harmless, and honest.
Bias (in AI)
Systematic errors in AI outputs reflecting prejudices in training data. Can manifest as gender stereotyping, racial assumptions, or cultural insensitivity in generated content.
Constitutional AI
A training approach where AI models are given a set of principles (a 'constitution') and learn to self-critique and revise their outputs to comply with those principles. Reduces reliance on human feedback for safety alignment.