The AI writing glossary, in plain English

AI detection and humanizing come with a vocabulary that is half statistics and half marketing. This glossary defines the terms you will meet on detector reports, in university policies and on tool websites, in language you can repeat to a lecturer or a client without checking.

Each entry gives a short definition, a longer explanation with an example, why the term matters for writing that reads human, and how it relates to what Humanize 360 measures or changes. Where a term is used loosely by vendors, we say so.

Start with perplexity and burstiness if you want to understand detectors, with false positive if you have been flagged, and with Human Score if you want to know what our own number means.

Glossary

• Perplexity: what it means and why detectors care
• Burstiness: the sentence rhythm that gives writers away
• AI detection: what it is, how it works and what it cannot tell you
• AI humanizer: what the tool does and what it does not
• Paraphrasing: what it means and how it differs from humanizing
• Plagiarism vs AI detection: copied from where, versus written by what
• False positive: when a detector flags writing a human actually wrote
• Watermarking: how AI vendors propose to mark their own text
• LLM: what a large language model is and why its writing has a texture
• Token: the unit language models actually count
• Temperature: the dial that makes AI text safe or surprising
• Hallucination: when a model states something that is not true
• Prompt: the instruction that shapes what a model writes
• Zero-shot: asking a model to do something it has never been shown
• Fine-tuning: teaching an existing model a narrower job
• GPTZero score: what the number means and what it does not
• Human Score: how our 0 to 100 detector scale is built
• Originality.ai score: what the percentage means, as far as anyone outside knows
• Stylometry: measuring a writer's fingerprint in numbers
• N-gram: the word runs behind stock-phrase detection
• Readability grade: what the number measures and what it misses
• Register: matching the formality of your writing to its reader
• Contraction rate: the small habit that separates registers

How the entries are grouped

• Detection signals: perplexity, burstiness, stylometry, n-gram, contraction rate, readability grade, register.
• Detector outputs: AI detection, false positive, GPTZero score, Originality score, Human Score, watermarking.
• Model terms: LLM, token, temperature, hallucination, prompt, zero-shot, fine-tuning.
• Writing practice: AI humanizer, paraphrasing, plagiarism versus AI detection.

Common questions

Are these the official definitions?

They are accurate working definitions written for writers, not for researchers. Where a term has a precise technical meaning we give it, then explain what it means in practice.

Can I cite this glossary?

You can, but for coursework you should cite a primary source. Each entry names the concept clearly enough to find one.

Why does an AI humanizer publish a glossary?

Because our detector shows eleven named signals and we want people to understand them. A number without an explanation is not useful to anyone.