What the published research actually says about AI text detection, and what to do when a score has real consequences for someone.
A practical, ordered guide for students and writers: what evidence to preserve first, what to say, what the research supports, and what not to do.
Two example emails, one if you wrote it yourself and one if you used AI, and the principles behind a good first reply.
Where to find version history in Google Docs, Word and Pages, what else to collect, and the habits that keep a record of your writing process.
The documents to ask for, the one-page timeline to build, how to respond to a detector score, and what to do afterwards.
Deadlines, the grounds panels accept, how to structure the appeal, and the argument to make when the case rested on a detector.
For students who did use AI: the first 24 hours, how to be honest without making it worse, and what to bring to the meeting.
Why a false confession and a false denial both backfire, and how to answer when you are not sure your tool use was allowed.
From warnings to expulsion: the usual penalties, what goes on your record, and knock-on effects for aid, visas and applications.
How to find the rule for your course, which uses are usually fine or usually not, and how to protect yourself in the grey area.
Why editing tools can raise a detector score, which features count as generative AI, and what to show if you are flagged.
These tools measure register, not authorship. The published false-positive figures, and the four reasons the problem does not get engineered away.
Perplexity, burstiness, trained classifiers and watermarking — what each method measures and the specific way each one breaks.
The base-rate arithmetic behind detector accuracy claims: why even a 1% error rate leaves a large share of flags pointing at people who did nothing.
Why second-language writing is flagged far more often, what the Stanford study found, and practical steps for writers and the people assessing them.
The false-positive arithmetic at institutional scale, why the errors land on specific students, and assessment approaches that do not depend on detection.
How to find out what happened, how to work with the school, and what changes when your child is at college.
For freelancers, employees and applicants: checking what was agreed, showing your process, and handling payment disputes.
Text-only AI detection does not work reliably, and the reason is structural rather than a temporary engineering gap. Almost every statistical marker anyone has found for machine writing turns out to be a marker of register — how formal, even and conventional the prose is — rather than of who produced it. That means the human writing most likely to be flagged is the most careful: scientific abstracts, formal essays, and work by people writing in a second language.
Three findings worth carrying around:
That includes the tool on this site. The analyzer on the home page has no validated accuracy rate and should not be used as evidence about anyone. It exists to make these statistics visible, not to settle questions.