AI Detection Glossary
AI detection relies on technical concepts from natural language processing, information theory, and machine learning. This glossary explains the key terms in plain English — what they mean, how detectors use them, and why they matter for anyone working with AI-generated text.
AI Detection
Software that identifies whether text was generated by an artificial intelligence model like ChatGPT, Claude, or Gemini. AI detectors analyze statistical patterns in writing — including perplexity, burstiness, and token probabilities — to classify text as human or AI-written.
Read full definition →AI Humanizer
A tool that rewrites AI-generated text to make it undetectable by AI detection software. Unlike simple paraphrasers, true AI humanizers target the statistical patterns — perplexity, burstiness, and token probability — that detectors analyze.
Read full definition →Burstiness
The variation in sentence length and complexity throughout a piece of writing. Human text is naturally bursty — mixing short, punchy sentences with long, complex ones. AI text tends to be uniform, with sentences of similar length and structure.
Read full definition →Entropy (in AI Detection)
Shannon entropy measures the randomness or information content in text. In AI detection, low entropy indicates predictable, formulaic writing typical of language models, while higher entropy suggests the varied word choices characteristic of human authorship.
Read full definition →False Positive (AI Detection)
When an AI detector incorrectly flags human-written text as AI-generated. False positives are a significant concern in academic settings, where they can lead to wrongful accusations of academic dishonesty against students who wrote their work entirely by hand.
Read full definition →GLTR
Giant Language Model Test Room — a visual analysis tool developed by MIT and Harvard that highlights text based on how predictable each word is. Words in green are highly probable (AI-like), while yellow, red, and purple indicate increasingly surprising choices.
Read full definition →GPTZero
An AI detection tool created by Edward Tian at Princeton University in 2023. GPTZero pioneered the use of perplexity and burstiness as the two core signals for identifying AI-generated text, and remains one of the most widely used standalone detectors.
Read full definition →Paraphrasing vs Humanizing
Paraphrasing rewrites text using different words and sentence structures while preserving meaning. Humanizing goes further by specifically targeting the statistical patterns — perplexity, burstiness, token probability — that AI detectors measure. Paraphrasing alone does not bypass modern AI detection.
Read full definition →Perplexity
A measure of how predictable the next word in a sequence is. Low perplexity means the text follows expected patterns closely, which is a hallmark of AI-generated writing. Human writing typically has higher, more variable perplexity.
Read full definition →Text Classifier
A machine learning model that categorizes text into predefined groups. In AI detection, text classifiers are trained to distinguish between human-written and AI-generated text by analyzing statistical features like perplexity, burstiness, and token distributions.
Read full definition →Token-Type Ratio (TTR)
The ratio of unique words (types) to total words (tokens) in a text. A lower TTR indicates more word repetition. AI-generated text often has a distinctive TTR pattern — moderately diverse vocabulary used with unnaturally consistent distribution across a document.
Read full definition →Turnitin AI Detection
Turnitin's built-in AI detection classifier, integrated into its plagiarism detection platform and used by over 16,000 institutions worldwide. It analyzes text at the sentence level and reports a percentage-based AI score alongside traditional originality checks.
Read full definition →