Form AI-1 (Rev. 2026)

Office of Academic Integrity

Preliminary finding

Suspected unauthorised use of generative artificial intelligence

Case no.
Opened
Examiner
Automated
Appeal
See §5

1

Used only to print on this page, in your browser. It is not sent anywhere and not stored.

2

Nothing you paste here leaves this device. This page makes no network request at all once it has loaded, contacts no third party, sets no cookie and runs no analytics. The fonts are served from this domain for the same reason. Open your network tab and watch. Every detector that charges you for this keeps what you give it.

§4

What this costs when someone means it

61.3%

of TOEFL essays written by Chinese students were falsely flagged as AI-generated by detectors. The same study put US-student essays at 5.2%.

Liang et al., Patterns, 2023 — Stanford

The axis is care, not authorship

On this page, a caricature of assistant prose scores 8 of 12 and an unrevised restaurant review scores 3. So the indicators do track something, but what they track is how formal and careful the writing is. Students taught to write formally, and people writing in a second language, are pushed up the same axis as the machine. That is the mechanism behind the number above.

The em dash never was the tell

Models overuse it because the best writers did first. As Mustafa Ocal of Florida International University put it, writers objecting to the accusation are saying I've always used em dashes. AI copied me, not the other way around, and, he adds, they're right. Reporting on The Economist's analysis found a lack of punctuation to be the likelier signal, with only one major model exceeding human em-dash rates.

Institutions are backing away

More than twenty-five universities have banned or restricted AI detectors, among them MIT, Yale, NYU, UC Berkeley, Vanderbilt, UCLA, Toronto, UBC and Manchester. Not because cheating stopped, but because the tools could not tell them who had done it.

It is not only students

Nike was publicly accused of using AI over a single em dash in a post about the tennis player Jannik Sinner. Every writer who has ever leaned on the mark now writes under quiet suspicion, and a good many have started avoiding it — which is a real change to how people write, produced by a rule that does not work.

§5

Right of appeal

If you have been accused

Ask which detector was used and what score it returned. Ask what the tool's own vendor documentation says about false positives — most state plainly that output should not be used as sole evidence. Ask whether a human read the work. Offer your drafts, your version history, your notes. Point to the Stanford figure above if English is not your first language.

If you are about to accuse someone

A detector score is not evidence. It is a number produced by measuring formality, and this page exists to show you that measuring formality convicts Emily Dickinson. Talk to the person. Ask them about their argument, their sources, their process. If they wrote it, they can tell you how. That conversation is the only instrument here that works.

What this page is, plainly

It is not an AI detector and must never be used as one. It returns no probability, no percentage of machine authorship, and nothing you could paste into a disciplinary email. It measures twelve folk rules, honestly, in order to show you what they are worth.

The Office of Academic Integrity above is invented. There is no office, no examiner and no case. The figures underneath it are real, sourced and linked.

Nothing you typed left your device. Nothing was stored, no cookie was set, no analytics run here, and the page makes no request to any third party — not even for its own typefaces.

Form AI-1 (Rev. 2026) · a companion piece to undercoveraiagents.site, which does the opposite job: catching the machines that really are pretending.