AI Detection Glossary
Clear definitions for reading detection results and discussions about generated writing with more care.
AI Detector
A tool that examines features of a passage and estimates whether they resemble patterns associated with generated writing. It is a signal for review, not proof of authorship or intent.
AI Humanizer
Software that rewrites text. Rewriting can change tone and rhythm, but it cannot prove the origin of a passage or make a claim about who wrote it. Review the output for accuracy, voice, and sources.
Burstiness
A descriptive term for variation in sentence length and structure across a passage. It can be one input in a model, but it is not a reliable human-versus-AI test by itself.
ChatGPT
A conversational AI product from OpenAI built on the GPT family of large language models. It generates human-like text in response to prompts and is one of the most common sources of AI-generated content that detectors are asked to identify.
E-E-A-T
Experience, Expertise, Authoritativeness, and Trustworthiness. It is a framework discussed in Google Search guidance for assessing content quality. It is not a detector score or a shortcut to a search ranking.
False Positive
When an AI detector incorrectly flags genuinely human-written text as AI-generated. Minimizing false positives is critical in academic and professional settings, where a wrong flag can have serious consequences.
GPT
Short for Generative Pre-trained Transformer, the model architecture behind ChatGPT and many other AI writing tools. GPT models are trained on large text corpora to predict the next most likely word in a sequence.
Hallucination
When an AI model generates information that is fluent and confident but factually incorrect or entirely fabricated. Hallucinations are a key reason AI-generated content should be verified before publication.
Large Language Model (LLM)
An AI model trained on vast amounts of text to understand and generate human language. Examples include OpenAI GPT, Google Gemini, Anthropic Claude, and Meta Llama. LLMs power most modern AI writing tools.
Natural Language Processing (NLP)
The field of artificial intelligence focused on enabling computers to understand, interpret, and generate human language. AI detection relies on NLP techniques to analyze writing patterns.
Perplexity
A measure related to how predictable text is to a language model. Predictability varies with topic, editing, language, and style, so it should be treated as one model feature rather than a verdict about an author.
Prompt
The instruction or input given to an AI model to produce a response. The wording of a prompt strongly influences the style and content of the generated text.
Tokenization
The process of breaking text into smaller units called tokens, often words or word fragments, so a language model can process it. Different systems can divide the same passage differently.
Training Data
The collection of text used to teach an AI model language patterns. The scope and quality of training data shape how a model writes and, in turn, the patterns detectors look for.
Transformer
A neural network architecture, introduced in 2017, that uses attention mechanisms to model relationships between words. Transformers are the foundation of modern large language models.
Watermarking
A technique intended to embed a detectable pattern in generated output. Its usefulness depends on whether the generating system supports it and whether later edits preserve the pattern.
Zero-Shot Detection
A model making a prediction about a class it was not directly trained to recognize. It may generalize to new systems, but that generalization should be evaluated instead of assumed.
Use a result as a starting point
Run a check, read the passages in context, and decide what deserves another look.
Try the AI Detector