Defense File

Build your AI-accusation defense file.

Five questions, about ninety seconds. You get an evidence checklist, a drafted response letter with the citations for your detector already in it, and the lines to avoid — yours to download, copy or print. Nothing is sent anywhere; everything runs on this page.

What to do if you are accused of using AI

The hours right after the email are the ones that matter, and almost everything that goes wrong in these cases goes wrong there. Four things, in this order: screenshot the accusation before anything is edited or withdrawn; preserve your writing history — version history, drafts, notes, sources — without touching the original file; ask in writing for the specific evidence being relied on; and only then reply, with the record in hand. What you should not do is apologise, offer an explanation you have not evidenced yet, or accept an informal resolution to make the discomfort stop.

The builder above assembles all of that into one file for your particular situation. If you would rather read first, the 72-hour playbook covers the same ground in long form, and how to prove you didn't use AI goes through evidence collection tool by tool.

Why the detector score is not the end of the conversation

A flag is not a finding. Independent peer-reviewed research reports false-positive rates of 5 to 20 percent across commercial detectors, and Stanford researchers found 61.3 percent on essays by non-native English writers. Turnitin's own documentation admits about 4 percent at the sentence level and describes the tool as advisory. Vanderbilt University disabled Turnitin's AI detector in August 2023 after calculating that roughly 750 of its 75,000 annual submissions could have been wrongly labelled. OpenAI discontinued its own classifier in July 2023, citing low accuracy. Your defense file quotes the numbers for the detector named in your case — the full table is on are AI detectors accurate.

What the file contains

  1. Evidence to gather tonight — a checklist built for how you actually wrote the paper.
  2. A drafted response letter — with the verified citations for your named detector already in it, and brackets for the facts only you have.
  3. The four-element structure behind that letter, so you can rewrite it in your own voice.
  4. Notes for your case — jurisdiction, detector, and the factors you flagged (ESL, accommodation, visa status).
  5. What not to say, matched to the stage you are at.
  6. Free help organisations for your country.

It is free, it takes about ninety seconds, and nothing you enter leaves your browser. If you are accused of using AI and did not, the file is the fastest way to walk into that meeting with a record instead of a denial. Everything here is informational and is not legal advice — see the disclaimer.

Frequently asked

I was flagged for AI but didn't use it — what do I do first?

Preserve the writing history before you reply to anyone.

Screenshot the accusation in full, with the sender and timestamp visible. Then open your document's version history and export or screenshot it — and stop editing that file, because further edits change the record you need. Collect drafts, notes, sources and any messages where you discussed the assignment. Only then reply, and reply by asking in writing for the specific evidence being used against you. The builder above turns this into a checklist for your writing tool and your stage of the process.

What exactly do I get at the end?

A checklist, a drafted letter, the notes for your case, and what not to say.

Six sections: (1) the evidence to gather tonight, specific to how you wrote the paper; (2) a drafted response letter with the citations for your named detector already inserted; (3) the four-element structure behind that letter, so you can rewrite it in your own voice; (4) notes for your jurisdiction, your detector and any factors you flagged; (5) the phrasings to avoid at your stage of the process; (6) the free help organisations for your country. You can download it as a text file, copy the letter, or print it.

Is anything I type sent to a server?

No. It runs in your browser and stores nothing.

The builder is a deterministic form, not an AI model. It runs entirely in your browser, sends no data anywhere, stores nothing, and sets no tracking cookie for your answers. Refreshing or closing the page discards everything. That is deliberate: nobody preparing a misconduct defense should have to hand their case to a third party first.

Should I send the draft letter exactly as written?

No — edit it. It's a starting structure, not a template to paste.

Every bracket in the draft is something only you can fill in, and panels read differently when the words are the student's own. Use the structure, keep the citations, and rewrite the prose so it sounds like you. A letter that reads as copied from a website undercuts the point you are making, which is that you write your own work.

Is this legal advice?

No.

It is informational, and it is built to prepare you for a conversation with your institution — or with a licensed attorney if the stakes justify one. If your visa, scholarship, enrolment or a professional licence is at risk, see your rights and when to retain an attorney, and speak to a licensed practitioner before you file anything.