Is using AI for homework cheating? What schools actually say in 2026

The short answer

Using AI for homework is cheating when the AI does the cognitive work the assignment existed to create. It is not cheating when AI helps your child do that work themselves.

That distinction is clean in principle and messy in practice, which is why schools are struggling and families are getting blindsided.

Why school policies are not helping

In practice this means the same behavior can be fine in one classroom and a discipline referral in the next. Your child is expected to intuit a boundary the adults have not agreed on.

The single most useful thing you can do this week: email your child's teacher and ask, in writing, what AI use is permitted on their assignments. It takes two minutes and it converts a guess into a documented answer.

Where most teachers actually draw the line

Across the policies that do exist, a rough consensus shows up.

• Generally allowed: explaining a concept, generating practice questions, checking grammar on the student's own writing, brainstorming topics.
• Usually a grey area: outlining an essay, summarizing a source, debugging code, translating.
• Almost always prohibited: generating submitted text, solving assigned problems, writing code that is graded as the student's own.

Notice the pattern — the line tracks whether the AI output ends up in the submitted work, and whether the assignment was testing the thing the AI did.

Suppose an assignment is intended to practice reading a difficult article. Asking for a summary could remove the very practice the teacher wanted, even if the final response is written independently. Ask whether the task is assessing comprehension of the reading, the student's argument, or both. Permission to check spelling on a draft doesn't imply permission to skip the source material.

Coding offers another concrete distinction. Imagine your child has written a loop that never stops. Asking why its condition remains true can support a debugging lesson where that help is permitted. Requesting an entire finished program avoids choosing the logic in the first place. Keep the original code and ask your child to explain the correction. The class policy still decides whether either form of help is allowed; professional habits aren't automatically classroom permission.

The false accusation problem

AI detectors do not detect AI. They score how statistically predictable text is, then infer. That inference fails in both directions: AI text edited lightly slips through, and human text written in a careful, formal register gets flagged. We break down the numbers in are AI detectors accurate?.

Students most at risk of a false positive are the ones writing in a second language, students who have just been coached to write more formally, and students with plain, direct writing styles.

How to protect your child before it happens

• Have them draft in a tool with version history (Google Docs) rather than pasting a finished block of text.
• Keep outlines, notes, and scratch paper for major assignments.
• For big projects, have them write a one-line note on what AI was used for, even if nobody asked.

If your child is accused

Ask what specific evidence the accusation rests on. If the answer is a detector score, say plainly that detector scores are probabilistic and known to produce false positives, and offer the process evidence instead — version history, drafts, and your child explaining the work out loud. Most cases resolve there.

Ask for the policy in writing, too. If it turns out the school had no clear published rule at the time, that matters.

Imagine an email says an essay was flagged by a detector. Before replying in anger, make a short timeline with your child: when they read the sources, drafted the outline, and revised the paragraphs. Gather the drafts that actually exist, along with the assignment instructions. Don't manufacture missing evidence or alter a document to make its history look different. A clear account of the process is more useful than a confident denial unsupported by details.

For the meeting, choose one passage your child can discuss in ordinary language. Ask how they selected the evidence and whether anyone or any tool helped change the wording. If they are anxious about speaking, ask the school how they can explain their process fairly. A difficult conversation is not a reliable memory test, and an inability to remember one sentence doesn't settle authorship.

If your child did use help that wasn't allowed, separate the admission from the plan for repair. Ask the teacher which work must be redone and what assistance is permitted during that redo. At home, identify the point where the shortcut began: misunderstanding the prompt, running out of time, or believing disclosure wasn't necessary. A smaller first step next time, such as a handwritten claim and two supporting notes, can make the boundary easier to follow. None of these preparations guarantees a particular school decision, but they keep the discussion grounded in the actual work.

The better long-term move

Detection is a losing arms race. Proof of learning is not. If your child can consistently explain their work, and you have something concrete showing they understood it, the question of what tool touched the document stops being the interesting one.

For the full framework — the permission ladder, the explain-it-back test, and how to set rules per assignment — read our parent's guide to AI homework help, then print the house rules.