An email arrives from your child's school. The teacher writes that your child has been flagged for using AI on an assignment they say they wrote themselves. The teacher believes the detector. You don't know who to believe.
This is happening to more families than schools admit, because AI detectors are wrong far more often than they claim. A flag is not proof. It is a statistical guess by a tool that misfires often.
What's really happening
A 2023 Stanford study published in the journal Patterns tested seven of the most widely used AI detectors. It found they falsely flagged 61% of essays by non-native English speakers as AI-written. The same bias affects students who write formulaically — including many neurodivergent students. The senior author advised institutions to avoid the tools entirely.
[SECTION IMAGE — delete this line and insert the "what's really happening" image here]
Universities have responded. Vanderbilt disabled Turnitin's AI detector in August 2023 after calculating that even Turnitin's own 1% false-positive rate meant about 750 of their students could have been wrongly flagged in a single year. OpenAI shut down their own AI classifier the month before; their own evaluation found it caught only 26% of AI-written text while wrongly flagging 9% of human writing.
When a school flags a student, they are trusting a tool the makers of AI themselves could not make work.
Whether your child has already been accused or you want to be prepared, do this tonight.
Pick one recent piece of writing. Sit with your child and ask specific process questions:
- What made you pick this thesis?
- Where did you find this source?
- Why did you open this way?
- What was the hardest part to write?
- What changed between your first draft and the final?
Listen. Take notes. If your child can answer specifically — naming sources, describing struggles, defending choices — the work is theirs, and your notes are evidence.
If you have been having the regular conversations recommended in the Thinking article, you already have months of this evidence built up. The same parental practice that catches outsourced thinking also documents authentic process.
If you are facing an accusation right now, also gather Google Docs version history, browser history showing research, time-stamped notes, and draft files. If your child uses a study or writing tool that logs activity — Mindgrasp auto-saves study sessions to a dashboard, and Lex.page keeps version history alongside a toggle that shows exactly which text was written by AI versus the student — pull those records too. Bring everything to the meeting.
When this isn't enough
Two harder cases:
If the school stands behind the AI flag despite your evidence
Ask which specific tool was used and what its documented false-positive rate is. Cite the Stanford research and Vanderbilt's decision to disable Turnitin. Request a meeting with the administrator above the teacher. If your child is a non-native English speaker, neurodivergent, or writes formulaically, mention that — the bias is documented.
If your child did use AI — in a way you believe was appropriate
If your child used AI for grammar checks, for exploring different ways to explain a concept, for finding sources, or for getting feedback on a thesis — but did the thinking and writing themselves — the case is different. The argument is not that AI was never used. It is that AI was used as a tool, the way a professional writer or researcher or many working teachers themselves use it. A blanket prohibition on AI prepares students poorly for the workplaces they are entering. Ask the school what specifically concerns them: the act of using AI, or the substitution of AI for the student's thinking. Those are not the same problem.
Next steps
Families need tools that prove authentic learning, not detectors that guess. A handful of platforms have started to provide this kind of evidence — Mindgrasp logs study sessions, and Lex.page tracks writing version by version with a toggle showing AI-written versus student-written text. Each helps for the slice of work it covers.
A system like Blast Learning is built around session-level records that span the full learning lifecycle, so the session history itself becomes documentation.
The bottom line
An AI detector's flag is a guess, not a verdict. What makes your child's work defensible is their ability to explain it — and that's something you start building tonight, one conversation at a time.