Oct 2, 2026

Suspect a Student Used AI? A Grade 6-12 Guide to the Evidence, the Conversation and the Call Home

What to check before you say a word, what to say to a 7th grader versus a junior, how to write the email home, and why a detector score should never decide it.

Flint title card on a cream brushstroke background: Blog 2026, Suspect a student used AI? What to do next

You're working through a stack of essays and one stops you. The vocabulary has jumped two grade levels since September, the paragraphs are tidier than this student's ever are, and none of it sounds like the kid in the third row. You have a hunch. The researchers at Harvard's Center for Digital Thriving call that hunch "AI-dar."

Plenty of teachers are where you are. When the center asked educators to describe an AI dilemma in their own words, 74% of the teachers who answered named something about cheating. Its report, published September 30, also found that suspicion now runs in both directions. Teachers can't prove a student used AI, students can't prove they didn't, and some students turn in weaker work on purpose so they won't be flagged. One 16-year-old sophomore called it "kind of like a hellscape for both students and teachers."

This guide is for grade 6-12 teachers holding a piece of work that feels off. The short version: gather evidence of how the work was made before you say anything, talk with the student privately and start with questions, base your decision on what they can show and explain, and tell the family what you saw rather than what you suspect. Never act on a detector score alone.

What should I do first if I suspect a student used AI?

Say nothing yet. Spend ten minutes collecting evidence of how the work was made, because "it doesn't sound like her" won't hold up when you talk to the student, a parent or an administrator.

You can check most of it without the student:

  1. Put the piece next to something the student wrote in class, by hand or on a timed prompt. A big gap in voice, vocabulary or structure is worth writing down. On its own it proves nothing, since students do have good weeks.
  2. If they wrote in Google Docs, open File > Version history > See version history. Google's help page explains how it shows who changed the file and when. An essay that grew over three evenings looks very different from one that arrived in two big chunks. You need edit access to the file to see it.
  3. Pull out any notes, outlines or drafts you collected along the way.
  4. If the work happened in a classroom tool that records paste history, check it. Flint is one of these (more at the end).
  5. Reread what you told the class about AI for this assignment. If the rule was fuzzy, the student may not know they crossed it. In an NPR/Ipsos poll of 545 K-12 teachers this spring, only 33% said their school had a formal policy on student AI use.
  6. Check whether the student has an IEP or 504 plan that mentions AI or assistive writing tools (see below).

Then write down exactly what you noticed. "Three paragraphs appeared in the document within one minute at 11:52 p.m." is something you can discuss with a 13-year-old and their parents. A feeling about tone isn't.

Should I use an AI detector?

If your school gives you one, never act on a score alone. Treat a flag as a reason to look at the student's process, and let the process decide.

The detector makers say the same. Turnitin's guide to its AI writing report warns that the model "may misidentify human-written, AI-generated, and AI-paraphrased text," which is why "it should not be used as the sole basis for adverse actions against a student."

The mistakes don't fall evenly, either. In a 2023 Stanford study, seven detectors were "near-perfect" on essays by U.S. eighth graders but, on average, labeled 61% of essays by non-native English speakers as AI-generated. In a 2024 Common Sense Media survey, 20% of Black teens said a teacher had flagged their schoolwork as AI-generated when it wasn't, compared with 7% of White teens and 10% of Latino teens.

Flint goes further than "use it carefully." In 2023 we committed to never building or selling an AI writing detector, and every step in this guide works without one. Philip Vinogradov, director of innovation, teaching and learning at Merion Mercy Academy, described how the school handles it in our case study: "AI detectors are not really reliable, so we often resolve those issues in conversation with the student."

How do I talk to a student about suspected AI use?

Meet privately, keep it short, and open with "Walk me through how you wrote this." A student who did the work can usually tell you where an idea came from, what they cut and which part gave them trouble. A student who didn't will often struggle with the same questions, and that tells you more than a percentage could.

A few rules hold at every grade:

  • Ask about the work, not about AI. "How did you come up with this example?" gets a real answer. "Did you use ChatGPT?" gets a yes or no, and usually a no.
  • Have the piece and your notes in front of you, and point to specific sentences.
  • Don't decide in the meeting. "Thanks, I'm going to think about this and get back to you tomorrow" keeps you from reacting to a shrug or to tears.
  • Give the student a way to show what they know: rewrite one paragraph in class, explain a choice out loud, or answer a question about the source.

Beck Tench, a co-founder of the Center for Digital Thriving, told EdWeek that students hold back their effort "in environments where accusations don't get any kind of conversation." The conversation is the part that keeps honest students from deciding it's safer to write badly.

What to say to a middle schooler (grades 6-8)

Keep it concrete and low-stakes. An 11- or 12-year-old may not know that what they did counts as cheating, and many will freeze if the meeting feels like they're in trouble.

"I want to hear how you put this together. Tell me about the part where you wrote [sentence]. What made you think of that?"

"Show me where you started. Did you jot notes first, or start typing right away?"

"This word here is a tricky one. What does it mean in your own words?"

If it becomes clear they used AI: "Thanks for being honest with me. Having AI write it isn't allowed on this one, because the point was for you to practice [skill]. Let's work out how you'll redo it."

At this age, a supported redo usually teaches more than a zero. Contact the family early, because middle schoolers rarely sort this out alone.

What to say to a high schooler (grades 9-12)

Older students can handle a more direct conversation, and they understand the stakes. Say specifically what you saw, then give them room to explain.

"When I compared this with your in-class essay, the writing was very different, and the version history shows most of it appeared at once. Help me understand how you wrote it."

"Talk me through your argument without looking at the paper. What's your strongest piece of evidence?"

"If you used any tools, which ones, and for which part?"

If they deny it and you still have doubts: "Okay. I'd like you to write a response to a similar prompt in class on Thursday so I can see how you work. That helps me grade this fairly."

For juniors and seniors, a grade can touch transcripts, recommendation letters and college applications. Follow your school's academic integrity process to the letter and keep a written record of each step.

What should I tell the student's parents?

Tell them what you saw, not what you suspect, and invite them in. Most published advice on suspected AI use is written for college instructors, who rarely deal with families. In grades 6-12, how the first contact with home goes often decides how the rest goes, and families of middle schoolers especially should hear from you before a grade shows up in the portal.

Here's an email you can adapt:

Subject: [Student]'s [assignment name]

Hi [Parent name],

I'm writing about [Student]'s [assignment], turned in on [date]. Parts of it read very differently from the writing [Student] has done in class, and the document's version history shows most of the text was added in one sitting. I met with [Student] today to talk about how it was written.

Before I decide anything about the grade, I'd like [Student] to [write a short response in class / talk me through the essay]. I'd also value your perspective, and I'm happy to talk by phone this week.

For this assignment, our class guideline on AI was: [one-line summary or link].

Thank you, [Your name]

On the phone, keep the same order: what you noticed, what you've already done, what happens next, then a question for them, such as "Has [Student] mentioned this assignment at home?" Some parents will arrive defensive, especially if they've read about students being wrongly accused. Our guide for parents tells them to ask what the concern is based on, so have your evidence ready to share.

When should I bring in the school?

Involve your department head or an administrator when the work is a major grade, when it's a repeat incident, or whenever your school's integrity policy requires it. Loop in the counselor if the student seems overwhelmed. A student who turns to AI at midnight before a deadline is sometimes struggling with more than the assignment.

If your school has no student AI policy, write down what you told the class about AI for this assignment and when, and ask an administrator how they want cases handled before you set a consequence. Our AI policy library collects policies from schools that have written one, including how they approach academic integrity without detectors.

What if the student has an IEP or 504 plan that allows AI?

Talk to the student's case manager before anything else. Some plans allow speech-to-text, word prediction or AI writing support, and whether something counts as misuse depends on how the accommodation is worded. If the plan doesn't address it, ask the case manager how it applies to this assignment and use the answer to set expectations for the next one.

Does this work for math, science and other subjects?

The steps are the same, but the evidence looks different outside an essay.

  • In math, the worked steps are the process. Apps can solve a problem from a photo and show every step, so have the student solve a similar problem at the board, or explain why they took one particular step. Students who did the work can usually find their own mistake when you point to the line.
  • In science, look at the lab notebook and the raw data. A conclusion that doesn't match the numbers the student recorded is a better thing to ask about than a polished paragraph.
  • In computer science, ask the student to change their code while you watch, like adding a small feature or fixing a bug you introduce.
  • In world languages, a two-minute spoken follow-up in the target language shows what the student can actually produce.

How do I keep honest students from feeling suspected?

Tell students how you'll check their work before anyone is accused, and check everyone the same way. When the process is known, your strong writers don't have to wonder whether a good essay will be held against them.

  • At the start of each assignment, say which AI uses are allowed and how you'll look at their process: drafts, version history, a short conference.
  • Collect process from the whole class, like an outline or a draft checkpoint, so asking for it never singles anyone out.
  • Don't treat polished writing, a big vocabulary or a particular punctuation mark as evidence. Some students write well.
  • Comment on growth you can see between drafts, so students know you notice the effort.

What should I change in the next assignment?

Build the process into the assignment so the evidence exists before anyone is suspected. A brainstorm in class, a draft checkpoint, a two-sentence reflection on which tools the student used and how, or a five-minute conference all do this. Start the writing in class when you can, and say plainly which parts AI may help with.

FAQ

Can you prove a student used AI?

Rarely with certainty. What you can usually find out is whether the student can explain the work and do something similar in front of you, which is what matters for grading. Version history, drafts, an in-class sample and a conversation tell you far more together than any score.

Are AI detectors accurate?

Not accurate enough to act on by themselves. Turnitin says its own AI writing report may not always be accurate and should not be the sole basis for action against a student, and a 2023 Stanford study found detectors wrongly flagged most essays written by non-native English speakers.

Should I give a zero for suspected AI use?

Follow your school's policy, and don't grade on suspicion. Many teachers have the student redo all or part of the work in class, which gives a grade you can stand behind and gives the student a fair chance to show what they know.

What if a student denies using AI?

Don't argue the point. Ask the student to show their process or complete a similar task in class, and decide based on what you see. If you still have concerns, follow your school's academic integrity process and bring in an administrator.

How do I check if a student used AI in Google Docs?

Open the document and choose File, then Version history, then See version history. It shows who changed the file and when. Large blocks of text that appear all at once are a reason to ask questions, though some students draft somewhere else and paste in their own work.

How Flint shows teachers the process

I work at Flint, so weigh this section accordingly. Flint is an AI learning platform for schools. Students use it through activities their teachers create, and teachers can open any student's session in those activities and read the whole conversation with Sparky, Flint's AI tutor.

Teachers can also see paste history. Hover over a student's message in a session and click the clock icon, and Flint shows any text the student pasted into the chat box while writing that message, along with when they pasted it. The text is saved as it was pasted, so it's still there if the student reworded it before sending. Our help center has the steps.

Vinnie Vrotny, director of technology at The Kinkaid School, described what that visibility is for in a 2024 case study: teachers can "see where a student may have made a leap of learning and had an epiphany moment within their learning versus when they've taken the shortcut and just copied and pasted the response back without participating in it."

Paste history has limits. It only covers Flint's chat box, so it won't show work done in another tab, text a student retyped, or anything written in Google Docs. It shows what was pasted, not where it came from. Treat it like every other piece of evidence in this guide: as a reason to ask the student a question.

Try it below. The students and sessions are made up.

Try it · Example sessions

What paste history shows a teacher

Three made-up students in three classes. Click the clock on any student message to see what they pasted.

Back to classPriya S.Argument paragraph: later start times Only visible to you
can you check my paragraph? later start times are a big deal for teens well-being. research shows teenagers body clocks shift so early mornings are really hard. so later start times could help grades and mental health.
Sparky
Thanks for sharing it. Before I give feedback: which research are you thinking of, and how does it connect to students at our school?
idk i just think kids are tired first period
Sparky
That's a real observation, and it could be your evidence. What have you seen in first period that shows it?
half my bio class has their heads down at 7:30

This is a session as the teacher sees it. Click the clock on a student's message to open its paste history.

1 of 4