Falsely Accused of Using AI? Here’s What to Do Next

Key Pointers

• Being wrongly flagged is more common than most students realise. In survey data, slightly more students reported being accused when they had not used AI than when they had.
• A detection score is a probability estimate, not evidence of authorship. That distinction changes how you respond.
• Your strongest defence is documentation you already have: version history, drafts, research notes, browser history.
• Ask what the accusation rests on before you argue. If it rests on a detector percentage alone, many institutions have already restricted that use.
• A sample appeal request is included below. Keep it factual, because the tone of your first response shapes the whole process.

The Short Version

If a professor has accused you of using AI on work you wrote yourself, you have more options than the situation feels like it offers. Detection scores are probabilistic, documented to be biased against certain writing styles, and increasingly restricted as sole evidence. Your response should be evidence rather than argument: gather your version history, request the specific basis in writing, and respond calmly using the template below.

First, understand what actually happened

An AI detector does not know who wrote your paper. It scores how closely your text’s statistical patterns match what language models produce, mostly through perplexity (how predictable your word choices are) and burstiness (how much sentence length varies).

Neither has anything to do with authorship. They describe the shape of your prose, which means a student who writes in a clear, consistent, well-edited style can score high while having written every word.

Our breakdown of AI detector false positives covers the mechanism, and our piece on why human writing gets flagged covers the characteristics that raise risk.

How often students are falsely accused of using AI

If this has happened to you, you are not an outlier.

In the Quetext Writing Integrity Survey, more than four in ten students reported being suspected or accused of using AI. The composition is the striking part: around 22% said they were accused when they had not used AI, against around 20% accused when they had. Among those accused, more than half maintain it was wrong.

Educators corroborate this. One in five said outright they had wrongly suspected a student, and a further third said they were not sure, which amounts to much the same admission.

The research explains why. Stanford HAI documented that AI detectors are biased against non-native English writers, flagging their work at disproportionately high rates. Sadasivan et al.’s 2023 paper on the reliability of AI-text detection found no detection method holds up reliably across adversarial conditions.

None of this proves your individual case. It does establish that a flag alone is weak evidence, which is the argument you will be making.

My professor accused me of using AI: what to do first

The instinct is to reply immediately and emotionally. Resist that. Work through these in order.

Step 1: Do not admit to anything you did not do. Students under pressure sometimes accept a lesser penalty to end the discomfort. If you wrote the work, saying otherwise creates a record that is hard to undo.

Step 2: Secure your evidence first, because some platforms overwrite history. Export or screenshot your Google Docs version history, Word track changes, or editor revision log. Save research notes, browser history from the period, annotated PDFs, and any messages discussing the assignment.

Step 3: Request the specific basis in writing. Ask which tool was used, what score it returned, which passages were flagged, and whether the detector output is the only evidence. The answer shapes your response.

Step 4: Check your institution’s policy. Many universities have restricted detector-driven accusations in the past year. If yours states detection cannot serve as sole evidence, that is directly relevant.

Step 5: Respond in writing, factually. Use the template below. Keep it short and unemotional.

If you want to know what a detector actually sees in your writing before that conversation, running it through Quetext yourself takes a minute.

What a false positive actually means technically

A detector returns a probability, not a verdict. A score of 85% does not mean 85% of your paper was AI-written, nor that there is an 85% chance you cheated. It means the text’s statistical patterns resemble model-generated writing to that degree.

Several things push human writing into that range: writing in English as a second language, formal academic prose, technical or formulaic fields like lab reports, careful self-editing that removes the variation detectors read as human, and short passages with too little text for a stable reading.

If any describe you, say so specifically. “I am a non-native English speaker and this bias is documented” is stronger than “the detector is wrong.” Our analysis of what the accuracy data shows covers the numbers if you need something concrete to cite.

Sample appeal request template

Adapt this to your situation. Keep it to one page and attach your evidence.

Dear Professor [Name],

Thank you for raising your concern about [Assignment Name], submitted on [Date]. I want to address it directly: I wrote this work myself, and I would like to provide evidence of my process.

I have attached:

1. Version history showing the document drafted over [number] sessions between [date] and [date]
2. My research notes and the sources I consulted
3. [Any other material, such as outline drafts or library records]

So that I can respond fully, could you confirm:

1. Which detection tool was used and what score it returned
2. Which specific passages were flagged
3. Whether the detection output is the sole basis for this concern

I would note that AI detection tools return probability estimates rather than determinations of authorship, and published research documents elevated false positive rates for [relevant factor: non-native English writing / formal academic prose / technical writing]. [University Name]’s policy on [cite if applicable] addresses this.

I am happy to discuss the paper in person at any time, including my argument, sources, and structural choices.

Thank you for considering this.

[Your name]

Three notes. Offering to discuss the content is powerful because a student who wrote the paper can defend its reasoning. Asking for the specific basis is often productive, since instructors sometimes realise the evidence is thinner than it felt. And citing your institution’s own policy carries more weight than citing research.

If your case escalates, our guide on how to write an appeal letter for plagiarism covers the longer-form structure.

If the first response does not resolve it

Most cases end after a documented, calm reply. If yours does not, escalate in order.

Ask about the formal appeals process. Most institutions have one, and using it creates a record.

Involve your academic advisor or student union. Advisors know local norms, and unions often provide representation in misconduct hearings.

Request a second reader. Asking for someone other than the accusing instructor to review the work sometimes resolves things quickly.

Keep every communication. Save emails and note details of verbal conversations immediately.

Watch the timeline. Appeals have deadlines, and reporting has documented students left waiting months on unresolved AI appeals.

NBC News reported on students now using AI tools defensively, specifically to avoid being flagged for work they wrote themselves. A fairly clear signal the current system produces perverse incentives.

How to protect yourself going forward

Draft where version history is recorded. Google Docs and Word keep it automatically. Do not compose in a notes app and paste the finished text, because the paste erases your evidence.

Keep your research trail. Saved sources, annotated readings, and notes corroborate your process.

Check your own work before submitting. You learn where you stand privately, and can look at a flagging passage in advance rather than being surprised.

Try this: Run your own paper through Quetext’s AI Detector before you submit. You get a probability score with sentence-level highlights showing exactly which passages might draw attention. The first 1,000 words are free.

The honest bottom line

We build an AI detector, so take this with appropriate scepticism: a detection score should never be the sole basis for a misconduct finding.

It is a probability estimate on text patterns, it carries documented bias, and the cost of acting on a wrong one falls entirely on a student who cannot prove a negative. Used well, a scan tells an instructor which passages deserve a closer read. Used as proof, it produces exactly the situation you are in. That distinction is your argument, and it is a legitimate one.

Check where your writing stands with Quetext before anyone else does.

FAQs

What should I do if my professor accused me of using AI?

Secure your evidence before responding, since some platforms overwrite version history. Export your document revision log, research notes, and browser history. Then request in writing which tool was used, what score it returned, which passages were flagged, and whether detection output is the only evidence. Respond factually, offer to discuss the paper’s content, and avoid admitting to anything you did not do.

• Save version history and research notes immediately
• Ask in writing for the specific basis of the accusation
• Offer to discuss the content, which a genuine author can do

How do I prove I did not use AI to write my paper?

The version history is such important evidence, since Google Docs and Word record the way a document is constructed gradually through drafts and changes, thus making it hard to fake. Adding to that evidence with notes made during research and saved links makes argument construction even more convincing.

• Version history showing how the document has been constructed
• Research notes and saved links
• Knowledge of the subject and successful conversation

Can AI detectors be wrong about student work?

As a regular thing, Detectors identify patterns like predictability of words and variability of the sentences, rather than the authorship. Hence, an academic writing that is formal, technical, edited and written by non-native speakers may be given high scores. A study from Stanford indicated that there is bias against non-native English writers, and surveys indicated that many students confess that they have been wrongfully accused as opposed to being correctly accused.

• Detectors identify patterns in language, not authors
• There is evidence of bias against non-native English speakers
• More people claim that they are wrongfully accused than being correctly accused

Is a high AI score proof that I cheated?

No. A score of 85% does not imply that AI wrote 85% of your paper or that you had an 85% chance to pay heed to it. A score tells us that your writing matches AI-created text to that extent in terms of statistics. There is no reliable method of detection, as published research indicates. Thus, in many universities, the use of a detector’s output is no longer enough in misconduct cases.

• Score gives percent of probability rather than conclusion
• A score does not mean that 85% of the paper was AI written
• Many universities do not allow to use a detector’s output as the only proof in case of dishonesty

What if my appeal gets rejected or ignored?

Proceed through your school’s formal appeals channels; doing so will create a written record. If you have an academic adviser or a union, they can provide assistance and representation at disciplinary hearings. If possible, request that the work be reviewed by someone other than the accusing instructor. Write down all communications as well as the details of any verbal conversations that you had immediately.

• Use formal procedures to create an official record of your case
• Seek help from your adviser or from the student union to get representation in the hearing
• Memorize all communications and keep track of the deadlines