A detector flag is a probability, not proof
Start here, because it changes everything about how you respond. An AI detector does not "catch" AI — it outputs a likelihood score based on statistical patterns, and those scores are far less reliable than the headline numbers suggest. Independent testing has put some detectors' real-world accuracy as low as 22–40%, and a widely cited Stanford study found detectors wrongly flagged around 61% of essays by non-native English speakers as AI-written.
This is now well enough established that institutions are backing away from these tools: the University of Waterloo discontinued Turnitin's AI detector in September 2025, and Curtin University followed in January 2026, both citing unreliability and the risk of wrongly accusing students. In July 2026, HEPI described AI detection in UK universities as a "fairness crisis." So if the case against you rests on a detector percentage, that percentage is contestable — and you are not the first person to contest it.
A number from an AI detector is evidence to be weighed, not a verdict. The burden is on the university to show misconduct on the balance of probabilities — a single tool's score usually isn't enough on its own.
Stay calm and read your university's process
An allegation is not a finding. UK universities have a formal academic misconduct procedure, and you have rights within it: to be told clearly what you're accused of and which policy applies, to see the evidence, and to a genuine chance to respond before any decision is made.
So before you reply to anything:
- Find your university's academic integrity / misconduct policy (usually on your student intranet) and read the procedure and possible outcomes.
- Contact your students' union advice service. Most SUs have trained advisers who support students through misconduct cases for free, and can attend meetings with you.
- Don't reply in a panic. A rushed, emotional email can be misread. You're allowed to take a little time to prepare a considered response.
Your students' union adviser, personal tutor and student support services exist for exactly this. Being accused is stressful — lean on them, and look after yourself while it's ongoing.
Gather your evidence
The single most powerful thing you can show is your writing process — because a genuine piece of work leaves a trail, and generated text usually doesn't. Pull together whatever you have:
- Version history. Google Docs and Microsoft Word both keep revision history showing your document evolving over hours and days. This is often the most persuasive single piece of evidence.
- Drafts and notes. Earlier versions, outlines, handwritten planning, annotated reading — anything that shows the thinking behind the essay.
- Sources. Your actual reading, highlighted PDFs, library loans, browser history around your research.
- Timestamps. When files were created and last edited, email trails with supervisors, submission logs.
If you can talk a marker through how you built the argument — why you cut a paragraph, where a source changed your mind — you're demonstrating the one thing an AI can't fake: your own reasoning.
How to respond to the allegation
When you put your response together, keep it factual and calm:
- State plainly that the work is your own and that you did not use AI to generate it (or, if you used AI in a permitted way — say, to check grammar — say exactly how, honestly).
- Present your process evidence — version history, drafts, notes — as your main argument.
- Note the limits of AI detectors factually: they produce probability scores, not proof, and have documented false-positive rates, especially for non-native English and neurodivergent writers.
- Follow the procedure exactly — deadlines, forms, who to send it to — and keep copies of everything.
This is general information, not legal advice, and every university's process differs. For serious cases or appeals, your students' union or a qualified adviser can help you prepare formal representations.
Why this keeps happening
It helps to know you're not imagining the unfairness. Detectors are trained to spot "average", predictable prose — which means clear, well-structured, grammatically clean writing can look "AI" to them. Students who write in a tidy, formulaic way, who use tools like Grammarly heavily, who are neurodivergent, or who speak English as an additional language are all flagged more often, through no fault of their own. The tool's weakness becomes the student's problem. That's precisely why sector bodies are now urging universities not to rely on detector scores as evidence.
How to protect yourself next time
You can't control a detector, but you can make your own process easy to show — and see problems before a marker does. A few habits help:
- Write in a tool that keeps version history (Google Docs, Word) and don't paste a finished essay in from elsewhere in one go — let the document show its working.
- Keep your drafts and notes until your mark is confirmed.
- Check how your own writing reads before you submit. That's what SafeGrade is for — paste your essay in and you get a private read on your writing patterns and an AI-risk indicator, so if your genuine work happens to read as "AI", you find out while you can still add a note to your marker or flag it up front. It's a review tool for your own peace of mind — not a way to game anything — and your essay stays on UK servers, never shared with your university or Turnitin.
Being accused when you did nothing wrong is not something you should have to manage. But knowing how detectors actually work, keeping your process, and checking your own writing first put you in the strongest possible position if it ever happens.
reads — before anyone else does.