What To Do If You Are Accused of Using AI in Your Dissertation: An Honest Opinion by a Dissertation Expert

Being a dissertation expert from last 12 years, I have seen this problem appearing again and again since 2022.

Quick answer: Being accused of using AI in dissertation work does not mean you are guilty or that your degree is finished. Stop editing the document, pull every draft and version history you have, and prepare evidence of your writing process before your meeting. The rest of this article shows you exactly how, for both US and UK processes.

I have been reading dissertations for 12 years now. Editing them, defending them, sitting in on vivas, fixing the stats chapters at 2 am before a submission deadline. In the last two years, I have had more emails opening with “I have been accused of using AI in my dissertation” than in the previous ten years combined.

So let me be direct with you. Getting accused of using AI in your thesis does not mean you are guilty. It means you now have a process to go through, and that process rewards people who stay calm and organised.

Knowing your dissertation expectations upfront, what the panel is actually judging, how long this drags on, how much evidence is enough, makes the whole thing far less frightening. This article tells you exactly what that process looks like, in the US and in the UK, and what I would do if I were sitting across the table from you right now.

Why AI Accusations Are Happening to Honest Students

Universities did not suddenly get better at spotting AI writing. They bought detection software and started trusting the score more than they should.

How Turnitin, GPTZero, and Copyleaks Actually Score Your Writing

Turnitin, the tool most institutions use, states in its own AI writing detection FAQ that it targets a document-level false positive rate under 1 percent. That sounds reassuring until you look at the sentence level. Independent detector-testing groups have measured Turnitin’s per-sentence false positive rate closer to 4 percent, rising further for second-language writers. One document has hundreds of sentences, so that gap matters more than the headline number suggests.

GPTZero and Copyleaks are the two other tools you will hear about most. GPTZero has publicly disputed at least one third-party ranking that scored it poorly, reporting its own false positive rate at around 0.05 percent after correcting what it says was a flawed test methodology. Copyleaks claims a false positive rate closer to 0.2 percent. My honest take: every one of these numbers comes from the vendor itself or from a testing group with its own methodology, so treat all of them as claims to verify, not settled fact, when you are building your defence.

A widely cited 2023 study by Weber-Wulff and colleagues tested 14 AI detection tools, Turnitin and GPTZero included, and found none of them scored above 80 percent accuracy overall. That single study is the reason several universities stopped trusting detector scores as standalone proof.

Definition: An AI detector false positive is when a detection tool like Turnitin, GPTZero, or Copyleaks incorrectly labels human-written text as AI-generated, usually because the writing is formulaic, simple in sentence structure, or written by a non-native English speaker.

Why Universities Are Already Backing Away From These Tools

This is not a fringe concern. Vanderbilt University disabled Turnitin’s AI detector entirely in 2023, and several other US universities followed within months. If a university itself does not fully trust the tool, you should not walk into your meeting assuming the panel will either.

First 24 Hours After Being Accused of Using AI in Dissertation Work

Do not panic-edit your document. Do not resubmit anything. Do not go quiet and hope it blows over. Here is the sequence I tell every client to follow.

1. Read the exact allegation. Was it a document-level AI score, a sentence-level flag, or a supervisor’s suspicion based on writing style change? These are three very different problems.
2. Stop touching the file. Any edit after the allegation looks like tampering, even if your intention is innocent.
3. Pull every draft you have. Old Word docs, Google Docs, email attachments to your supervisor, anything with a date on it.
4. Check your version history. Google Docs version history and Word’s track changes with timestamps are some of the strongest proof of authorship you can offer. It shows a writing process, not a paste job.
5. Gather your working files. If your dissertation involves statistics, your SPSS syntax files, R scripts, or Stata do-files carry their own timestamps and edit history. Most students forget this evidence, and it is often stronger than the writing itself because it shows your actual analytical thinking.
6. Write a timeline. When you started each chapter, when you met your supervisor, when you got feedback. Memory fades fast under stress, so do this on day one.

Do You Need a Lawyer? Can You Refuse the Meeting?

Most first-instance meetings do not require a lawyer, though some UK institutions allow a student union advisor or, for serious cases, external legal representation. Check your own institution’s policy rather than assuming either way. Refusing to attend is almost always the wrong move. Panels are entitled to proceed and decide in your absence, so a missed meeting usually makes your position weaker, not stronger.

How to Prove You Wrote Your Dissertation Yourself

This is the part most guidance articles get wrong. They tell you to “gather evidence” without telling you what actually counts.

Version History and Your Draft Trail

Version history is your strongest card. A document that grew over three months, with visible false starts, deleted paragraphs, and supervisor comments, does not look like something generated in one sitting. AI-generated text tends to appear whole, with no messy editing trail behind it. Your messiness is your alibi.

Supervisor Correspondence

Your supervisor’s email trail matters more than students think. If you sent drafts, asked questions, or received feedback on specific sections, that correspondence proves engagement with the actual content, not just submission of a finished file.

Statistical Software Evidence

If you are a stats-heavy dissertation candidate, your analysis files are gold. A syntax file in SPSS or a do-file in Stata records every command you ran, when you ran it, and how your analysis evolved. Nobody accuses you of asking AI to write regression output. If your written interpretation of that output matches your own analytical logic and errors, that consistency is powerful proof.

What About Grammarly and Paraphrasing Tools?

Grammarly and similar grammar checkers are not the same as AI text generators, but they can still trigger a flag because they rewrite sentence structure in ways that look statistically similar to AI paraphrasing. If you used Grammarly, keep its own edit history or version log if the tool offers one, and be ready to explain exactly which features you used. Spelling correction versus full tone rewrites are very different in a panel’s eyes.

Preparing for the Academic Integrity Meeting or Hearing

Terminology splits here. In the US, you are usually looking at an academic integrity board or an honor code hearing. In the UK, it is more often called a misconduct panel, sometimes preceded by an investigatory viva where you defend your thesis verbally. Dissertation candidate or thesis writer, the evidence file a thesis writing expert would tell you to build is the same on both sides of the Atlantic.

US vs UK: How the Process Actually Differs

Aspect	United States	United Kingdom
Process name	Academic integrity board or honor code hearing	Misconduct panel, sometimes with an investigatory viva
Standard applied	Varies by institution, often a preponderance-style standard	“Balance of probabilities”
Who decides	Faculty and student integrity board	Departmental panel, sometimes with a separate viva panel
External appeal body	Internal appeals committee, varies by state and institution	Internal appeal, then the Office of the Independent Adjudicator (OIA)
Typical appeal window	Varies by institution	Commonly 10 to 28 working days, check your own policy

Whatever it is called, the standard your panel usually applies is some version of “more likely than not,” not proof beyond doubt. That is a lower bar than criminal law, which is exactly why your evidence needs to be organised and specific, not emotional.

What to Bring to the Meeting

• Every draft version with dates
• Version history screenshots or exported logs
• Supervisor correspondence
• Your raw data and analysis files
• A written timeline of your writing process
• Notes on your own reading list and sources, showing you engaged with the literature

What to Expect If You Are Accused of Using AI in Your Thesis at Viva

If you are heading into a viva specifically to defend against an AI accusation, expect content questions, not just process questions. They will ask you to explain your own arguments, your methodology choices, and why you interpreted your results the way you did. If you genuinely wrote the work, this is where you win. You cannot fake understanding of your own analysis under real-time questioning.

A note on scope: this article reflects patterns I have seen across dozens of cases, not legal advice. For anything involving possible expulsion or a formal legal process, loop in your students’ union representative or a solicitor who specialises in education law before your meeting.

What Happens Next: Consequences and Appeals

Possible Outcomes

Outcomes vary widely. I have seen cases dismissed outright once version history was reviewed. I have also seen resubmission with capped marks, degree classification reductions, and, in rarer serious cases, suspension. The deciding factor is usually how strong your process evidence is, and separately, how well the panel itself has been trained to weigh a detector score against that evidence, which varies a lot between institutions.

Will This Show Up on Your Transcript?

A dismissed allegation typically leaves no mark at all. A finding of misconduct, even a minor one, is usually recorded internally and can appear on your academic record depending on your institution’s policy, though it does not always appear on the transcript a future employer sees. Ask your department directly what gets recorded, since this varies enough that guessing is not worth the risk.

Appeals: What Actually Works

Yes, you can appeal an AI detection accusation. Appeal windows I have seen range roughly from 10 to 28 working days, but this varies a lot by institution, so check your own policy immediately rather than assuming a timeline, and do not sit on a poor decision. Appeals succeed most often when the university relied on the detector score alone without giving you a real chance to present your own evidence, a pattern the OIA has flagged repeatedly in its published case reviews.

Case example (composite, details changed): A UK master’s student I supported had a dissertation chapter flagged at document level. Her writing was clear and formally structured because English was her third language. We pulled her Google Docs version history covering six weeks of edits and her supervisor’s email feedback on two earlier drafts. The panel reviewed both and dismissed the case at the first meeting.

Case example (composite, details changed): A US client had his methodology chapter flagged after a style change mid-document. He had written the first half exhausted at 3 am and revised it later after coffee, changing his sentence rhythm. His Word track-changes history showed the entire revision process. Case closed after one hearing, no penalty.

Case example (composite, details changed): A UK undergraduate was flagged after running his entire introduction through Grammarly’s tone rewrite feature. He kept Grammarly’s own change log and walked the panel through exactly which suggestions he had accepted versus rejected. The panel accepted that tone-editing software is not the same as AI generation and closed the case.

My Honest Opinion as a Dissertation Expert

If you came to this article looking for a shortcut, there is not one. But if you came here panicking about whether your degree is finished, it almost certainly is not, provided you treat this like the process it is, not a verdict that has already been decided.

I have watched genuinely innocent students lose these cases because they walked in with nothing but their word. I have watched equally innocent students win comfortably because they walked in with a folder. Be the second kind.

One worry I hear a lot: does getting outside help to prepare make you look guiltier? No. Panels see your evidence, not who helped you organise it. Bringing a well-built folder to a meeting is no different from bringing a lawyer’s letter to a contract dispute, it is just good preparation, not a confession.

If you want a dissertation coach or dissertation mentor to help you build that folder and prepare your response, that is exactly the kind of work I do with clients before hearings, not just before submission deadlines, including students hunting for a dissertation expert UK familiar with OIA appeals specifically. Plenty of people who reach out are not mid-crisis at all. Some just want an expert dissertation writer for a second pair of eyes on their chapters, or basic dissertation editing services near me to tighten the writing before resubmission. Both are smaller jobs than misconduct defence, and I handle both.

If you are still building your dissertation and want to avoid this entirely, read our guide on how to know if you need a dissertation expert before your next chapter, or look at how to organise your dissertation drafts properly so your version history works for you automatically. If your accusation touches your statistics chapter specifically, our statistical consulting services can help you reconstruct and explain your analysis clearly before any hearing. Students preparing for a viva in general will also find our dissertation defence timing guide useful, and if you want direct support right now, find a dissertation writing mentor here and get in touch before your meeting date.

FAQ

What evidence do I need if accused of using AI on my dissertation? Version history, supervisor email correspondence, dated drafts, and raw analysis files like SPSS or Stata syntax logs. Together these show a real writing and thinking process over time.

Can Turnitin’s AI detector be wrong? Yes. Turnitin’s own AI writing detection FAQ states a sub-1 percent false positive rate at document level, but independent per-sentence testing puts it closer to 4 percent, higher again for non-native English writers.

What should I bring to an academic integrity meeting? Bring drafts with dates, version history exports, supervisor correspondence, your data files, and a written timeline of when you wrote each section.

How do I prove authorship of my dissertation? Show your editing trail through Google Docs or Word version history, alongside supervisor feedback threads and your own analysis working files.

Can you appeal an AI detection accusation? Yes. Appeal windows vary by institution, commonly somewhere between 10 and 28 working days in the cases I have seen, so confirm your own deadline directly. Appeals often succeed when the decision relied on a detector score alone without reviewing your process evidence.

Do UK and US universities handle AI accusations differently? Broadly yes. US institutions typically run this through an academic integrity board or honor code hearing, while UK institutions use a misconduct panel, sometimes with an investigatory viva. See the comparison table above for the full breakdown.

Does Grammarly trigger AI detection flags? It can. Grammarly’s tone-rewrite and clarity features change sentence structure in ways that sometimes look statistically similar to AI paraphrasing, even though the tool is not an AI generator.

Is a Turnitin AI score proof of misconduct on its own? No. Turnitin itself states that its tool does not make a determination of misconduct. It provides a score for a human decision-maker to weigh alongside other evidence, not a verdict.

What happens if I’m found guilty of AI misconduct in my dissertation? Outcomes range from a warning and resubmission with capped marks to degree classification reduction or, in serious repeat cases, suspension. Very few first-time, well-evidenced cases end in expulsion.

Will an AI misconduct accusation show up on my transcript? A dismissed case usually leaves no mark. A confirmed finding is typically recorded internally and may or may not reach the transcript depending on your institution’s policy, so ask your department directly.

Author: Siddharth Gupta, LinkedIn

Siddharth has spent 12 years as a dissertation and statistics consultant, supporting students through misconduct hearings, vivas, and academic integrity boards across India, the US, UK, and UAE.