It's 11:48 pm. The essay is due at midnight. ChatGPT is one tab away.
You're not really asking whether you can use it. You're asking whether you'll get in trouble if you do.
The short answer: Using AI isn't automatically cheating. Turning in thinking you didn't do, when your class says the thinking has to be yours, is. Two things decide which side of the line you're on: your teacher's or professor's rules, and whether the AI helped you learn or did the learning for you.
Wait, is everyone using AI?
Pretty much. In a 2026 survey of 1,054 university students, 95% said they use AI in some form, and 94% said they use it on work that gets graded [1]. In a US survey of 1,047 college students, 97% said schools need to respond to AI cheating, and most wanted clear rules rather than detection software or outright bans [2].
So students aren't looking for a loophole. They're looking for the line.
Who actually decides what counts as cheating?
Not TikTok, and not the internet. Your teacher or professor does.
One class bans AI completely. The next lets you brainstorm with it. A third wants you to use it and show your prompts. All three can happen at the same school in the same week. Before you use AI on anything graded, run three quick checks:
- Find the AI rule. Look in the syllabus or the assignment instructions. If there isn't one, ask. A two-line email beats an academic integrity meeting every time.
- Check whether you have to say you used it. A lot of policies allow AI help as long as you disclose it.
- Ask what the assignment is testing. If it's testing your writing, AI writing it for you breaks the assignment. If it's testing your analysis, an AI grammar check probably doesn't.
What's the difference between getting help and cheating?
When the rules are fuzzy, use this test: after using the AI, could you do this yourself?
| Usually help (still check your rules) | Usually crosses the line |
|---|---|
| "Quiz me on my notes." | "Write the answer I'm turning in." |
| "Explain this concept a different way." | "Solve this problem set." (then you copy it) |
| "Here's my attempt. What's wrong with it?" | "Do it first, I'll read it after." |
| "Help me plan when to study for each test." | Submitting AI text as your own words without saying so |
The left side builds your skills. The right side replaces them. And as you'll see in a second, the right side tends to backfire even when nobody notices.
Can teachers tell if you use AI?
Sometimes. But not reliably, and that cuts both ways.
A Stanford study tested seven popular AI detectors on essays written by real people. Essays by non-native English speakers were wrongly flagged as AI 61% of the time on average, and about 20% were flagged by every single detector. Essays by native English speakers were almost never misclassified [3]. Detectors have improved since, and the companies dispute some of these findings. But the takeaway stands: a detector score is a clue, not proof.
Your best protection is a work trail. Write in apps that save version history, keep your rough drafts and notes, and if you used AI within the rules, keep the chat. If you're ever questioned, your process is your evidence.
Why "I didn't get caught" isn't actually a win
Here's the study that should make everyone pause.
Researchers followed nearly 1,000 high school math students. One group practised with a ChatGPT-style AI that gave answers. Another used a version that only gave hints. A third had no AI at all. Then everyone took an exam with no AI allowed [4].
Answer-giving AI: 48% better on practice problems. Then 17% worse on the exam than students with no AI.
Hint-giving AI: 127% better on practice. About the same as the no-AI group on the exam.
Read that again. Homework scores went up. Actual learning went down. Finals, the SAT, placement tests, job interviews: sooner or later, the AI isn't in the room with you.
Try this tonight
Before you open any AI on homework, give yourself ten honest minutes with the problem. Write a rough attempt, even a bad one. Then ask the AI, "What's wrong with my attempt?" instead of "Do this for me." Same tool, completely different result.
Where Crammm fits
We built Crammm for the left side of that table. You drop in your own class material, like your syllabus, notes, slides or even a photo of the whiteboard. Crammm turns it into quizzes, flashcards and out-loud practice where you answer in your own words. After every session it shows your Course Readiness, an honest read on what you actually know before the test finds out for you.
To be fair about it: Crammm can also summarise your notes, and a summary used instead of studying is still a shortcut. It also doesn't set your class's AI rules. Your teacher does. What it does is make the honest route the easy one.
Be one of the first to try it. Start free with 1,000 AI credits and a full day of unlimited access. Upload one class, run one session, and tell us what works and what breaks.
FAQ
Is it cheating to use AI to study?
Usually not. Using AI to quiz yourself or explain a concept is studying. Just check the rules for take-home tests and graded work.
Is Grammarly Considered AI?
It depends on your class. Many policies treat grammar and spelling fixes differently from AI that writes text for you. If you're unsure, ask before you submit.
What should I do if I'm falsely accused of using AI?
Stay calm, ask what evidence the concern is based on, and share your drafts, notes and version history. Ask how your school's appeal process works.
References
- HEPI and Kortext, Student Generative AI Survey 2026, 12 March 2026 (UK undergraduates; sponsored by Kortext)
- Inside Higher Ed Student Voice survey 2025, summary via Academica, 2 September 2025
- Liang et al., Stanford, GPT detectors are biased against non-native English writers, Patterns, July 2023, reported by The Markup, 14 August 2023
- Bastani et al., Generative AI without guardrails can harm learning, PNAS 2025, summary via Wharton
Take what helps. Adapt it to your courses. Build a routine that works for you.
← Back to all field notes