AI Detection Glossary — Key AI Detector Terms Defined
Plain-English definitions for AI detection score, false positive, perplexity, burstiness, and more — the vocabulary for reading any detection report, including ours, with a clearer eye.
Straight answers about AI detection, written the way we would want to read them: no scare tactics, no invented accuracy figures, just what the evidence actually shows and how to use it.
Plain-English definitions for AI detection score, false positive, perplexity, burstiness, and more — the vocabulary for reading any detection report, including ours, with a clearer eye.
A course AI policy is table stakes now, and "no AI allowed" is neither enforceable nor what most instructors actually mean. Here's how to write one specific enough to use.
A short, honest note about how you used a tool isn't a confession — it's what turns "did they hide something?" into "they showed their work." Here's how to write one.
272 words that have survived 160 years of scrutiny. We ran them — plus the Second Inaugural — through our own detector and published every number it returned, flagged sentences included.
Every AI detector claims to be the most accurate one. That's not a number you can verify, and it's the wrong thing to check first. Here's what actually tells you whether a detector is fair to use.
A detector can show you real evidence worth a conversation. What it can't do is stand in for one. Here's a fair process for using a detector with student work — from reading the evidence to talking with the student before any decision.
If careful, formal writing has ever made you wonder whether it reads as AI-generated, that worry is common and worth taking seriously — but guessing is a poor substitute for seeing the actual sentence-level evidence.
Formal register and a dense bibliography are exactly what trip up detectors that weren't built for academic writing. Here's what happens to a research paper or literature review in a scan built to handle both.
A new model name doesn't reset detection. Here's what AI-writing detectors actually key on across GPT-5, Gemini, and Claude — and what a result can and can't tell you.
A number on its own — 12%, 61%, 94% — isn't evidence. Here's what a detection score is actually measuring, why the same text can read differently on different tools, and what to look for instead of a bare percentage.
A humanizer tool can swap words and reorder sentences, but it's rewriting the surface of text whose underlying structure was already set. Here's what changes, what doesn't, and why laundering AI writing is a bad bet.
It isn't a clean yes or no. Here's what actually feeds a professor's impression — detection tools and the human eye test — and how to use AI honestly instead of hoping you don't get flagged.
A plagiarism checker and an AI checker aren't the same tool asking the same question. Here's what each one actually looks at in ChatGPT-style writing — and what neither one can prove.
Before you submit, you can see exactly what a pattern-based detector would see in your draft — and fix what's genuinely formulaic instead of guessing.