Your Kid Used ChatGPT for Homework. Now What?

The essay came back with a comment in the margin: “This doesn’t sound like your voice.” And then you found the browser history.

Or maybe you didn’t find it — you noticed that a piece of work that normally takes your 13-year-old 45 minutes to resist and then 40 minutes to write appeared, finished and submitted, in under 10 minutes. And the vocabulary was different.

Parents in this situation usually face two simultaneous questions: is this a disciplinary issue, and is this the end of learning? The research offers a more useful frame than either.

What the Research Actually Shows About AI and Cheating

Before anything else: the cheating-rate numbers are worth knowing, because they’re counterintuitive.

Academic dishonesty surveys conducted before ChatGPT’s release consistently found that 60–70% of students reported engaging in at least one dishonest academic behavior per month. A 2024 study in Computers and Education: AI (ScienceDirect) by Gao et al. comparing self-reported cheating behaviors before and after ChatGPT’s release found that the overall rate of cheating behavior had not meaningfully increased — and in some survey cohorts, slightly decreased.

The 74’s 2025 analysis of available research also found no consistent evidence of a cheating spike attributable to AI availability.

This is counterintuitive but explainable. Students who were going to cheat had a variety of methods before ChatGPT. AI shifted the method, not the underlying behavior. Meanwhile, students who weren’t going to cheat aren’t primarily using AI to generate dishonest work — they’re using it for research, drafting, editing, and idea generation in ways that are often appropriate.

By early 2025, research suggested that roughly 27–29% of high school students were using ChatGPT for homework, and 18–21% of middle schoolers. Most of this use is not straightforward academic dishonesty — it’s a spectrum ranging from appropriate research assistance to wholesale outsourcing of thinking.

The Real Problem: Learning Loss, Not Academic Integrity

The integrity question matters, but the learning question matters more. Here’s why:

Academic dishonesty is primarily a problem of performance — you’re misrepresenting what you know. That’s real, and it has real consequences if detected. But the deeper problem with AI-generated homework is what doesn’t happen: the struggle.

Robert Bjork’s research on desirable difficulties at UCLA is foundational here. The cognitive effort involved in working through a problem — searching for the right word, organizing an argument, figuring out why your paragraph doesn’t flow — is precisely the mechanism that builds durable learning. The mental work is the learning. When AI does that work, the student gets a grade without getting the underlying cognitive benefit.

A student who submits AI-generated work may pass the assignment. They will almost certainly struggle on the test, where there’s no AI to assist — not because they’re bad at the subject, but because they never processed the material.

This reframe is useful in the parent-child conversation: the problem isn’t primarily “you lied to your teacher.” The problem is “you didn’t learn anything, and that’s going to matter later.”

AI Homework Use: A Spectrum

Most parent and school responses treat AI homework use as binary — cheating or not cheating. The reality is a spectrum:

AI use type	Academic dishonesty?	Learning impact	Better alternative
AI writes entire essay, submitted as is	Yes	High risk — zero learning from task	Student writes draft, uses AI only for feedback
AI generates outline, student writes from it	Context-dependent	Moderate risk — structure outsourced	Student generates own outline, uses AI to evaluate it
AI explains a concept student doesn’t understand	Usually not	Minimal risk — AI used as tutor	Good use; teach student to verify AI explanation
AI gives feedback on a draft the student wrote	Usually not	Low risk — AI used as editor	Good use; discuss which feedback to take
AI used to research topic, student synthesizes	Usually not	Low risk — AI used as accelerant	Good use; teach source verification
Student submits AI work as own without disclosure	Yes	High risk	Have the conversation below

The position most schools are articulating for 2025–2026: AI as a thinking tool is generally appropriate; AI as a replacement for the student’s thinking is not. As of spring 2025, most schools had not communicated this clearly — 35% of district leaders reported providing student training on AI use; only 10% had formal written guidelines. Your child is likely navigating ambiguous norms.

The Three-Step Conversation

How you respond in the first 10 minutes matters. The goal is to open a conversation that’s productive, not to issue a verdict.

Step 1: Get curious before getting firm

“I noticed your essay didn’t sound like your usual writing. Can you tell me about how you wrote it?” leaves space for the conversation. “Did you use ChatGPT?” is a closed question that triggers a yes/no defense.

Children who confess use are usually showing you something worth knowing about their relationship to the task: it felt impossible, it felt pointless, they didn’t know where to start, the school’s AI policy was unclear, they were overwhelmed. Find out which one before moving to consequences.

Step 2: Separate the two issues clearly

“Here’s what I’m concerned about: one, whether this was honest with your teacher. Two, whether you actually learned what this was supposed to teach you.” Make both explicit. The first is about integrity. The second is about their future — what they’ll be able to do when AI isn’t available.

“Let’s think about both separately” is more productive than treating them as the same issue.

Step 3: Rebuild with AI, not without it

The answer to inappropriate AI use is almost never “no AI ever.” It’s teaching appropriate AI use. Ask your child to redo the assignment with AI open but with a different role: “Ask the AI to ask you questions about the topic, not to write it for you.” Or: “Write a draft, then use AI to tell you what’s weak about it.” The cognitive work stays with the student. The AI becomes a coach rather than a ghostwriter.

For a full framework on how to use AI as a thinking tool rather than an answer machine, see Teaching Kids to Use AI as a Thinking Partner. For the foundational AI concepts your child needs to use it responsibly, see What AI Literacy Means for a 10-Year-Old.

What to Do About the School Dimension

As of April 2026, only a minority of schools have written, student-facing AI policies. Many have contradictory policies where one teacher allows AI and another bans it with no coordination. Your child may genuinely not know what the current standard is.

The practical step: contact the teacher or school counselor to understand the specific policy before the incident becomes a formal disciplinary matter, if it hasn’t already. Come with information (“here’s what I found out from my child”) rather than defensiveness. Most teachers are navigating this in real time and appreciate a parent who is engaged rather than hostile.

What to Watch for Over the Next 3 Months

Week 2–3: After the conversation and any consequence, does your child’s subsequent work show evidence of engagement — their voice, their choices, their organizational quirks? Personal academic writing has characteristic patterns that AI doesn’t replicate. Changes in the texture of work are more telling than denials or confessions.

Month 2: Is your child now able to articulate what AI is good for in schoolwork vs. what it shouldn’t do? That distinction, internalized and stated back, means the conversation landed.

Month 3 self-check: Is your child’s AI use now visible and discussable in your household, rather than hidden? Hidden AI use persists because it’s shame-based. Open AI use that’s discussed and bounded is a much more sustainable position.

Frequently Asked Questions

Is it always cheating if my kid uses ChatGPT for homework?

No. The school’s AI policy determines what’s permissible, and most policies (for the minority of schools that have them) permit research assistance and drafting support while prohibiting wholesale generation and submission. In the absence of a clear school policy, the more useful question is whether your child is actually learning — if they’re using AI to accelerate their own thinking, that’s different from using it to replace it.

My child says everyone does it. Is that true?

Use is widespread — 27–29% of high schoolers by early 2025 — but “everyone submits AI-generated work unchanged” is not accurate. Many students use AI in ways that are appropriate or borderline, and a meaningful minority don’t use it at all. The “everyone does it” defense is worth engaging directly: “Some kids do, some don’t, and in three years when you’re expected to write something well, your grade on this assignment won’t matter — your ability will.”

Should I tell the teacher?

This depends on whether the assignment has already been submitted and graded. If yes, that ship has sailed; focus on what comes next at home. If the work is still upcoming, encourage your child to talk to the teacher directly about appropriate AI use for the assignment — which also models the kind of transparency that’s actually a valuable skill.

What if my child’s school has no AI policy at all?

Then your household policy fills the gap. Establish what appropriate AI use looks like in your home: AI as a coach and editor is fine; AI as the author is not. Put it in the family tech agreement (see The Mental Load of Tech Parenting for a framework) rather than relitigating it every assignment.

---

About the author

Ricky Flores is the founder of HIWVE Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.

Sources

1. Gao, J., et al. (2024). “Cheating in the age of generative AI: A high school survey study of cheating behaviors before and after the release of ChatGPT.” Computers and Education: AI, 6, 100175. https://doi.org/10.1016/j.caeai.2024.100175

2. The 74. (2025). “High School Cheating Increase from ChatGPT? Research Finds Not So Much.” https://www.the74million.org/article/high-school-cheating-increase-from-chatgpt-research-finds-not-so-much/

3. Stanford Graduate School of Education. “What Do AI Chatbots Really Mean for Students and Cheating?” https://ed.stanford.edu/news/what-do-ai-chatbots-really-mean-students-and-cheating

4. Nerdynav. (2025). “ChatGPT Cheating Statistics: Latest Facts on AI in Schools.” https://nerdynav.com/chatgpt-cheating-statistics/

5. Axios. (2025, August). “Confusing school policies on AI, ChatGPT use leave families guessing.” https://www.axios.com/2025/08/29/school-ai-policies-chatgpt

6. Bjork, R.A., & Bjork, E.L. (1992). “A new theory of disuse and an old theory of stimulus fluctuation.” In A. Healy, S. Kosslyn, & R. Shiffrin (Eds.), From Learning Processes to Cognitive Processes: Essays in Honor of William K. Estes, Vol. 2, pp. 35–67. Erlbaum.