Survive the SystemESSAYFree guide

AI Flag Appeal Playbook: How to Appeal an AI Detection Flag (Step-by-Step)

Build a proper evidence file, understand how the appeal process works at SA universities, prepare for a viva voce, and respond professionally — even if you're panicking right now.

By Akhil Boddu · AfricanTechBro29 April 2026

AI Flag Appeal Playbook: How to Appeal an AI Detection Flag (Step-by-Step)

What you'll be able to do after reading this: Build a proper evidence file, understand how the appeal process works at most SA universities, prepare for a viva voce, and respond professionally — even if you're panicking right now.

Prerequisites: Access to your university's academic integrity policy (usually on their website). A Google account (for Docs version history). Your assignment files.

Time to complete: 30 minutes to build your evidence file. The appeal process itself varies by institution.

If you're reading this, breathe

You got flagged. Maybe you used AI, maybe you didn't. Either way, what you do in the next 48 hours matters more than what happened before you submitted.

Most students panic and either ignore the notification or fire off an angry email. Both are mistakes. This guide walks you through exactly what to do, step by step, with the same advice I'd give a friend who just texted me "bro I got flagged."

The system isn't perfect. The detectors aren't perfect. But you need to play this smart.

What to do in the first 24 hours

Don't respond to anyone yet. Seriously. Not your lecturer, not the faculty office, not the academic integrity committee. You need information before you say anything.

Here's your first-day checklist:

Screenshot the notification and any AI detection report. Get the percentage flagged, which sections were highlighted, and any similarity scores. Save these somewhere that isn't your university email (in case you lose access).
Find your university's academic integrity policy. This is usually under "Academic Regulations" or "Student Conduct" on the university website. Some universities call it "Academic Honesty Policy" or "Assessment Integrity Framework."
Read the full policy. Not a skim — actually read it. You're looking for:
- What AI detection threshold triggers a formal investigation?
- What are the stated penalties for different levels of misconduct?
- What's the appeal process? How many days do you have to respond?
- Is there a distinction between first offence and repeat offence?
Save or print the policy. You'll reference specific clauses in your appeal.
Start collecting your evidence (covered in detail in Section 4 below). Don't wait.

⚠️ Some universities give you as little as 5 working days to respond. Missing the deadline can result in a default guilty finding. Check your deadline immediately.

Understanding the consequences

The range of outcomes is wide, and knowing where your situation falls helps you calibrate your response.

Typical penalty scale (from least to most severe):

Zero marks on the specific assignment
Fail the module (you retake it next semester/year)
Suspension for 1-5 years
Expulsion (permanent)

Most universities distinguish between a first offence and repeat offences. A first offence with a low AI percentage (say, under 20%) usually lands in the "zero for the assignment" range. A repeat offence or a fully AI-generated submission can trigger suspension or expulsion.

Some universities have zero-tolerance policies. UNISA, for example, treats any confirmed AI misconduct as grounds for suspension regardless of the percentage.

The severity usually depends on three things:

How much of the work was flagged
Whether this is your first offence
How you respond (cooperative vs. combative)

That third point is why this guide exists. Your response matters.

The burden of proof problem

Here's the part most students don't expect: at most universities, the burden is on you to prove you wrote the work. This is the opposite of how the legal system works ("innocent until proven guilty"). In academic misconduct proceedings, you're effectively guilty until you prove otherwise.

You will likely face a viva voce — an oral interview where a panel asks you to explain your work line by line.

The panel will ask questions like:

"Why did you choose this source?"
"How did you arrive at this argument?"
"Walk us through your research process for Section 3."
"What does this term mean in the context of your essay?"

If you can't answer these clearly, the panel assumes the work isn't yours. It doesn't matter what the AI detector said — your inability to explain your own work is the real evidence against you.

This is why Section 6 (viva voce prep) is so important. Read it even if you think your written appeal will be enough.

Building your evidence file

This is the most important section in this guide. Your evidence file is your defence. The stronger it is, the better your chances.

Google Docs version history

If you wrote in Google Docs: go to File, then Version history, then See version history. This shows every edit you made, with timestamps.

What you're looking for: screenshots showing gradual progression over multiple days. Real writing looks like incremental edits — a paragraph here, a revision there, sections moving around. AI-generated work appears as a single large paste event with minimal editing afterward.

Export screenshots that show:

The first version (your initial outline or rough notes)
Several intermediate versions showing edits over time
The final version

If you wrote in Microsoft Word 365, the same version history exists under File, then Info, then Version History.

Early drafts and outlines

Dig up anything from before the final submission:

Rough planning documents
Mind maps (photograph handwritten ones)
Bullet-point outlines
Notes from lectures that informed your essay
WhatsApp messages where you discussed the topic

Browser history

Your browser history shows your research trail. Export or screenshot the history for the days you were working on the assignment. Show the articles, journal papers, and websites you actually visited.

This is strong evidence because AI-generated work has no research trail. If you can show you spent 3 hours reading journal articles on the topic, that's hard to fake.

Chat logs

If you discussed the assignment topic with classmates in study groups, WhatsApp chats, or Discord servers — screenshot those conversations. They prove you were engaging with the material.

Grammarly or writing tool logs

If you used Grammarly, ProWritingAid, or similar tools, check their activity dashboard. These tools log every document you edited and when. This proves YOU were actively editing the text (not just pasting in AI output).

Why timestamps matter

The single most important thing across all your evidence: showing a process over time. AI-generated submissions typically appear as one event — a large block of text pasted into a document at 2am the night before the deadline, with no prior activity.

Real work shows days of incremental progress. That pattern is your best defence.

flowchart TD
    A["Get flagged"] --> B["Screenshot everything"]
    B --> C["Find university AI policy"]
    C --> D["Read the policy — note deadlines"]
    D --> E["Build evidence file"]
    E --> F{"Do you have version history?"}
    F -->|Yes| G["Export screenshots showing gradual edits"]
    F -->|No| H["Gather drafts, browser history, chat logs"]
    G --> I["Write your formal response"]
    H --> I
    I --> J{"Viva voce required?"}
    J -->|Yes| K["Prepare — re-read everything, practice explaining"]
    J -->|No| L["Submit written appeal with evidence"]
    K --> L
    L --> M{"Outcome?"}
    M -->|Cleared| N["Request written confirmation"]
    M -->|Denied| O["Appeal to higher body"]
    O --> L

The detectors are unreliable — use this in your appeal

This is where the facts are on your side. AI detectors have documented reliability problems, and several SA institutions have acknowledged this.

UCT scrapped Turnitin AI detection. In October 2025, the University of Cape Town officially stopped using Turnitin's AI detection feature because of unreliability concerns. If UCT — one of Africa's top universities — doesn't trust the tool, that's worth mentioning in your appeal.

UNISA's own data shows problems. UNISA reported an approximate 30% false positive rate in their disciplinary proceedings. That means roughly 1 in 3 students flagged by the detector didn't actually use AI.

Turnitin's own published rates. Turnitin says their false positive rate is "below 1%" at the document level, but around 4% at the sentence level. And those rates only apply to documents with 20% or more AI writing. For documents with less AI content, the error rates are higher and less well-documented.

Non-native English speakers get flagged more often. This is a documented issue across multiple studies. If English isn't your first language, your writing style may trigger AI detectors more frequently. Mention this if it applies to you.

Structured or formal writing triggers flags. Students who naturally write in a formal, well-organized style get flagged more often than those who write casually. If you're a strong academic writer, the detector may be penalizing you for being good at your job.

Include these facts in your written appeal. They don't prove you didn't use AI, but they establish reasonable doubt about the detector's accuracy.

How to prepare for the viva voce

The viva voce (oral interview/hearing) is where most students either save themselves or sink. Here's how to prepare.

The night before:

Re-read your entire assignment. Know it cold. You should be able to explain every paragraph without looking at the paper.

For each section of your work, be ready to answer:

Why did you include this section?
What sources informed this point?
How did you arrive at this conclusion?
What alternative arguments did you consider?

Practice out loud. Explaining something in your head is different from saying it. Practice walking through your methodology verbally. If you stumble or can't explain something, that's the section you need to study more.

Bring your evidence file. Print it. Organize it chronologically. Have it ready to hand over. Don't make the panel ask for evidence — offer it proactively.

During the hearing:

Stay calm and factual. The panel is looking for competence and honesty, not emotion.
Don't get defensive. If you're asked a tough question, take a breath and answer it directly.
If you genuinely don't remember why you phrased something a specific way, say "I don't recall my exact thought process for that sentence, but the argument I was making was..." Honest uncertainty is better than a fabricated explanation.
If you genuinely wrote the work, your confidence in your own knowledge is your strongest asset. People who wrote their own essays can talk about the content naturally. People who didn't tend to give vague, surface-level answers.

💡 The panel usually starts with easy questions ("Tell us about your essay topic") and gets more specific. Use the easy questions to build momentum and show you know the material.

If your appeal is denied

A denial isn't always the end. Most universities have multiple levels of appeal.

The typical escalation path:

Academic integrity committee (first hearing)
Faculty board or faculty appeal committee
University senate or central disciplinary committee

If your first appeal is denied:

Request the denial in writing with specific reasons. You need to know exactly why they ruled against you so you can address each point.
Address each stated reason with additional evidence in your next appeal.
Check the timeline. Most universities give you 10-15 working days to appeal to the next level. Don't miss it.
Consider getting advice. Some students in SA have consulted legal professionals — particularly for suspension or expulsion cases.
If you're cleared at any stage, request written confirmation that your academic record is clean. Don't assume it's been updated — verify it.

How to prevent this in the future

Whether or not you used AI this time, here's how to protect yourself going forward:

Always write in Google Docs or Word 365. Both auto-save version history. This is your automatic alibi.
Never copy-paste large blocks of text from anywhere — even from your own previous documents. Large paste events look suspicious in version history.
Keep all research notes and rough drafts. Even messy ones. A photo of a handwritten brainstorm on a napkin is evidence.
If you use AI for brainstorming or grammar checking (where your university allows it), document exactly how you used it. Screenshot the prompts. Save the chat. Transparency protects you.
Before submitting, run your work through an AI detector yourself. Tools like GPTZero, ZeroGPT, and Originality.ai give you a preview of what your university will see. If your score is high, you have time to revise before submitting.
Write your assignments over multiple sessions across multiple days. This creates a natural version history pattern that's clearly human.

Looking for Bluebook or help with a pre-submission check? Message me and tell me what you need. The old Bluebook link is currently unavailable.

Email template for your initial response

Subject: Response to AI Detection Flag — [Module Code] [Assignment Name]

Dear [Name/Title of the person who notified you],

Thank you for informing me about the AI detection flag on my [assignment name] for [module code], submitted on [date].

I take academic integrity seriously and would like to address this matter thoroughly. I did not use AI tools to generate the content of this assignment. [If you used AI for specific permitted purposes like grammar checking, state that here: "I did use Grammarly for grammar checking, which is permitted under the module guidelines."]

I would appreciate clarification on:

Which specific sections of my assignment were flagged
The AI detection percentage and which tool was used
The formal process for submitting my response and evidence

I have compiled an evidence file demonstrating my writing process, including version history, research notes, and drafts, which I am prepared to submit for review.

I am also available for a viva voce at your earliest convenience to discuss my work in detail.

Please let me know the deadline for my formal response and the next steps in this process.

Kind regards, [Your name] [Student number] [Contact details]

Evidence checklist

Summary

Don't respond immediately. Collect information and evidence first.
Your evidence file is your defence. Version history showing gradual edits over time is the strongest evidence you can have.
AI detectors have documented reliability problems. UCT dropped them. UNISA has a ~30% false positive rate. Use these facts.
Prepare for the viva voce. If you wrote the work, your ability to explain it is your best asset.
If denied, you can appeal to a higher body. Get the denial reasons in writing and address each one.
Going forward, always write in Google Docs or Word 365 and keep your drafts.

This guide powers the "Comment ESSAY" CTA on @africantechbro.ai