My child says they did not use AI. Should I believe them?

Most students flagged by detectors did not cheat. These tools measure how predictable writing is, so clear, careful, well-proofread work scores as machine-like regardless of who wrote it. Your child insisting they wrote it is entirely consistent with them having written it, and the first useful step is gathering evidence rather than interrogating them.

What should I ask the school?

Start with the central question: what evidence exists beyond the detector score? If the answer is only the score, that is the whole case. Then ask to see the written policy on how scores may be used, which tool was used and what it returned, whether the whole piece was flagged or part of it, and what the appeal route and deadline are. Ask for the outcome in writing.

What if my child did use AI?

Honest and specific beats vague and defensive. Many school policies permit grammar and language help while prohibiting content generation, so your child may be inside the rules without knowing it. Schools treat a discovered cover-up far more harshly than the original use, so a precise account now is both the honest route and the safer one.

Should I get a lawyer?

Rarely at the first stage, and threatening one in the opening email tends to harden positions. Most cases resolve at classroom or head-of-department level when a parent turns up with evidence and a reasonable manner. If the potential outcome includes exclusion, a permanent record, or a visa consequence, then advocacy is warranted — start with the school's own appeals process and any advocacy service available to you.

How do I prove my child wrote the essay?

Version history is the strongest evidence: if they drafted in Google Docs, screenshot the timeline showing the document being built over sessions. Add notes, printouts, library books, and browser history from the relevant evenings. Then ask your child to talk you through the essay — a child who wrote it can usually discuss it for ten minutes, which is exactly what a good teacher will test.

How do I stop this happening again?

Have them draft in Google Docs so version history is automatic, never paste a finished essay in from elsewhere, and keep their notes. It is five minutes of setup and it turns a future accusation into a screenshot rather than a hearing. Watch out for the opposite reaction too: some students start writing badly on purpose afterwards, which is a worse outcome than the accusation.

Your Child Has Been Accused of Using AI: A Parent's Guide

Q: Does it matter that English is my child's second language?

Considerably, and you should raise it explicitly. Published research found detectors misclassify non-native English writing at dramatically higher rates, because a smaller active vocabulary produces more predictable word choices. This is documented rather than anecdotal, and an assessor who understands the mechanism usually stops treating the number as evidence.

How to work out what happened, what to ask the school, and how to advocate well — including what to do if your child did use AI and is afraid to say so.

The email arrives on a Tuesday afternoon and your stomach drops. Your child's essay has been flagged as AI-generated, there is a meeting on Thursday, and you have no idea whether they did it.

This guide is for that week. It is written for parents and carers rather than for students, because your job is different from theirs — you have to find out what actually happened before you can help, and those two things need doing in the right order.

Before anything else: the conversation at home

Get this wrong and everything downstream gets harder.

Ask once, calmly, and mean it. Something like: "I'm not angry and I'm not going to be. I need to know what actually happened so I can help you properly. If you used it, tell me now — I can work with that. What I can't work with is finding out later."

Then stop talking and let the silence sit. Teenagers frequently confess into a pause and deny into a follow-up question.

Believe the answer, provisionally. Most flagged students did not cheat. Detectors produce false positives at meaningful rates, particularly on clear, careful, well-proofread writing. Your child insisting they wrote it is entirely consistent with them having written it.

Do not interrogate. Cross-examining a frightened teenager produces contradictions that mean nothing except that they were frightened. If they later face questions from the school, a rehearsed and anxious account is worse than a natural one.

Watch for the middle answer. The most common truthful answer is neither yes nor no. "I wrote it but I asked ChatGPT to fix the beginning." "I used it for the outline." "I wrote it in my own words but I used it to check my grammar." Those are all different situations with different consequences, and they are all defensible if disclosed now. They become indefensible if discovered later.

Then: gather evidence, do not edit anything

Before your child touches the file again.

The version history is the strongest thing you have. If they drafted in Google Docs, open File → Version history → See version history and screenshot the timeline. It shows the essay being built over time — paragraphs appearing, being deleted, coming back reworded. Generated text does not have that shape; it arrives whole.

If they wrote in Word on a laptop with no cloud sync, there is likely no history at all. That absence is not evidence against them, and it is worth saying so plainly at the meeting.

Collect the surrounding material. Notes, printouts, library books, browser history from the relevant evenings, messages where they sent a paragraph to a friend. Anything dated.

Ask them to talk you through the essay. Not to test them — to find out what they can demonstrate. Can they explain why they made that argument? What they cut? Which source they found least convincing? A child who wrote it can usually talk about it for ten minutes. This is also the best possible preparation for the meeting, because it is exactly what a good teacher will do.

What to ask the school

Go in asking questions rather than making accusations. You will get more, and you will not spend your credibility in the first five minutes.

Worth asking, in roughly this order:

What is the evidence beyond the detector score? This is the central question. If the answer is "the score," you have identified the whole case.
What does your policy say about detector scores? Many schools have adopted guidance stating that a score alone is not sufficient grounds. Ask to see the policy in writing.
Which tool was used, and what was the score? Different tools disagree substantially on the same text.
Was the whole essay flagged, or part of it?
Has my child's other work been compared? If this essay reads like everything else they have written, that is meaningful.
What is the process from here, and what are the possible outcomes?
What is the appeal route and the deadline? Note this even if you expect informal resolution.

Ask for the outcome in writing, whatever it is.

What is worth knowing about the technology

You do not need to become an expert, but three facts will change how the conversation goes.

Detectors do not detect AI. They estimate how statistically predictable writing is. Text that uses common words in conventional structures scores as machine-like — which describes clear, careful, well-edited student writing. A high score can mean your child wrote plainly and proofread thoroughly.

They cannot point to a source. A plagiarism checker shows you the document your child supposedly copied. An AI detector shows you nothing, because there is no source document. Any conclusion rests entirely on the number.

Vendors say so themselves. Turnitin's guidance to schools states its AI indicator should not be the sole basis for a misconduct finding. If the school is treating it as conclusive, they are going beyond what the supplier advises — and pointing that out politely is more effective than arguing about the science. There is more detail in what a Turnitin AI score actually means.

One group is affected far more than others. If English is your child's second language, this matters enormously. Published research found detectors misclassify writing by non-native English speakers at dramatically higher rates, because a smaller active vocabulary produces more predictable word choices. If that is your situation, raise it explicitly — it is documented rather than anecdotal. See why AI detectors fail non-native English speakers.

If your child did use AI

Different situation, and the instinct to protect them by denying it is the wrong one.

Honest and specific beats vague and defensive. "I used it to fix my grammar and rewrite two paragraphs" is a manageable conversation. A denial that collapses when someone asks a follow-up question is far more serious, and schools treat the dishonesty more harshly than the original use.

Find out what the policy actually permits. Many schools allow grammar and language support while prohibiting content generation. Your child may be inside the rules and not know it, or over a line they did not realise existed.

Separate the two failures. If they used it and then lied to you, deal with the lie at home and the use with the school. Merging them helps nobody.

Treat it as a teaching moment rather than a catastrophe. A teenager who used ChatGPT for homework has done something extremely normal for their generation and needs to learn where the line is — not to be told they are dishonest. The distinction that matters is between assistance with expression and substitution of thinking; we set it out in AI writing tools and academic integrity.

Advocating without making it worse

Be the reasonable one. Schools resolve things faster for parents who are calm and specific. Threatening lawyers in the first email hardens positions and rarely helps a first-instance case.

Focus on process, not the tool. "What evidence supports this beyond the score?" lands better than "your software is broken," even though the second may be true.

Bring the evidence, do not lecture. Two or three well-chosen exhibits — version history, notes, a draft — beat a printed stack of articles about detector failures.

Ask what would satisfy them. Frequently there is something simple: a conversation with the teacher, rewriting a section under supervision, walking through the sources. Offering makes you look confident, because you are.

Keep a record. Dates, who you spoke to, what was said. Follow up verbal conversations with a short email summarising them.

Escalate only if informal resolution genuinely fails. Most of these end at the classroom or head-of-department level. Save the formal appeal for when it is needed, and mind the deadline.

Looking after your child through it

The academic outcome usually matters less than people fear. The experience of being disbelieved often matters more.

Say explicitly that you believe them, if you do. Say explicitly that you will handle it either way, if you are not sure. Children flagged wrongly frequently describe the loneliness of being told a computer disagrees with them as the worst part — worse than the mark.

Watch for the aftermath, too. Some students start writing badly on purpose afterwards, adding errors and roughening their prose to avoid being flagged again. That is a bad outcome: it degrades their writing, wastes the effort they put into learning to write clearly, and teaches exactly the wrong lesson. If you see it, say so.

Afterwards

Whatever happens, set up habits that make this survivable next time. Drafting in Google Docs, keeping notes, and not pasting a finished essay in from somewhere else — the automatic version history has resolved more of these cases than any argument. The full list is in how to keep a writing audit trail.

It is unfair that this is necessary. It is also five minutes of setup, and it means a future accusation ends with a screenshot instead of a hearing.