Yes, AI can help a teenager study for exams, but only when it quizzes them instead of summarizing for them. The test is simple: at the end of a session, could your teen answer those questions with the chatbot closed? Study modes help, but a teen can switch them off in one tap, and your school's rules on AI always come first.
What does AI actually do well when a teen is revising?
It is good at the dull parts of preparing to study: turning a unit's notes into practice questions, asking them one at a time, and explaining a step the textbook skipped. It is poor at the part that makes revision work, which is the effort of pulling an answer out of your own head.
That distinction matters because the cognitive science on revision is older than any chatbot. A 2013 review in Psychological Science in the Public Interest by John Dunlosky and colleagues rated ten common study techniques. Practice testing and distributed practice earned the highest utility rating (the Association for Psychological Science's summary, January 2013). Summarization, highlighting and underlining, and rereading were among the techniques rated low. That review studied how people study, not AI, so what follows is my reading of it, not a finding about chatbots.
Here is my reading. Asking a chatbot to summarize your notes is rereading with a faster machine. Asking it to quiz you is practice testing. The same tool can do either, and a tired teenager at 10pm will drift toward the easy one.
Does the research show AI helps or hurts exam results?
It shows that how the tool is set up decides. The clearest study I know is a field experiment by Hamsa Bastani and colleagues, published in PNAS in 2025. It followed nearly 1,000 students in grades 9 to 11 at one high school in Turkey, studying math in the fall of 2023, across three conditions: textbooks and notes only, a plain chatbot, and a chatbot built to give hints instead of answers.
On the practice problems, the plain-chatbot group scored 48 percent higher than the textbook group and the hint-giving group scored 127 percent higher. On the later exam, taken without AI, the plain-chatbot group scored 17 percent lower than the textbook group. The hint-giving group scored about the same as the textbook group, which is to say the safeguards removed the harm but did not add a gain. PsyPost's summary of the study (April 2026) notes the limits: one school, one subject, a short period and short-term results.
That is evidence from one setting. My interpretation is that it describes a pattern every parent can check at home: the chatbot that helped most while it was open helped least once it was closed.
A good revision session ends with a closed laptop and a teenager who can still answer.
Do ChatGPT's study mode and Gemini's Guided Learning fix this?
They point in the right direction, and neither is a guarantee. This is what each company said when it launched the feature.
OpenAI's announcement of ChatGPT study mode (July 29, 2025) says it "helps you work through problems step by step instead of just getting an answer." It also says users can "easily toggle study mode on and off during a conversation," and that because it runs on custom instructions there may be "some inconsistent behavior and mistakes across conversations." The announcement mentions no parental control or school lock for it. Features change, so check ChatGPT's current settings rather than relying on a post written in November 2026.
Google's post on Guided Learning in Gemini (September 23, 2025) describes it as "an interactive study partner that helps you build a deeper understanding of any topic." The post does not state age limits or school-account rules for it. Our guide to Gemini's age and school-account rules covers what Google says elsewhere.
So what can a study mode not guarantee? That your teen leaves it on, that it never gives a wrong explanation, and that it matches your teen's syllabus or the exam board's marking. Treat it as a good default, not a safety setting. For a wider look at tutor-style tools, see AI tutors for kids: do they work, and which kind helps.
What I See in Schools
The pattern I see most often is that students revise the way they were never taught to: highlight, reread, recopy. When AI arrives, the same habit gets a new costume. "Make me a summary of chapter six" feels productive, produces something tidy, and tests nothing.
Teachers who get good results with AI revision tend to give students one rule: the tool asks, you answer. Questions come one at a time, the answer comes from the student first, and the explanation arrives only afterward. That is the school version of the research above, and it is a rule a family can copy tonight.
If your teen's school has a policy on AI for revision and coursework, ask what it says before you decide anything. Some schools allow AI for practice but forbid it for any work that is marked.
What can we do tonight?
Here is a 15-minute session that keeps the thinking with your teen. Sit nearby the first time, then step back.
- Pick one topic from a class, not a whole subject. A single unit is enough.
- Paste in the teacher's notes or the textbook pages, so the quiz matches the class and not the whole internet.
- Use a prompt like this: "Quiz me on this material, one question at a time. Wait for my answer before telling me if I am right. Do not give hints unless I ask. When I get one wrong, ask me a simpler version later."
- Do the whole session without looking at the notes.
- Close the laptop. Ask your teen to write down three things they got wrong, from memory.
- Two days later, repeat with the same topic. Spacing it out is the second half of what the research rated highly.
One more habit worth building: check anything surprising against the textbook or the class notes. In the Turkish study, PsyPost reports the plain chatbot gave correct answers only about half the time in the chat logs. Tools have changed since fall 2023, but the habit of checking has not become any less useful.
What can I say to my teen?
You could say: "I am not checking up on you. I want to see how you use it for revision, because the way you use it matters more than whether you do."
You could say: "Try the quiz version. Ask it to test you instead of explain things to you, and tell me tomorrow which felt harder."
You could say: "If you got every answer right and it felt easy, that is a sign to make the questions harder, not to stop."
If the conversation turns into an argument about cheating, how to talk to your teen about AI and homework has more words to try.
How does this change by age?
Under 8, exams are not the issue, and a chatbot is not needed. Quiz a child out loud, with a parent doing the asking.
At 8 to 12, a parent should do the quizzing, with paper cards or by reading questions aloud. If a chatbot is used at all, a parent sits beside the child and the tool's own age rule applies; our guide to the minimum age for each AI tool lists them.
At 13 to 15, use the quiz routine above with a parent nearby for the first sessions. Agree the rule together: the tool asks, your teen answers first.
At 16 to 18, your teen owns the process. Your job is the closed-laptop check, a weekly question about what they got wrong, and a standing offer to talk.
If exam stress is taking over your teen's evenings, talk to the school counselor, your pediatrician or your family doctor. A study routine does not replace that conversation.
When do the school's rules come first?
Always. A school can set its own rules on AI for homework, revision, coursework and assessed work, and exam boards and testing bodies set theirs. If you are unsure whether AI use for a particular assignment is allowed, ask the teacher in writing before your teen uses it. Ten questions to ask your child's school about AI is a good place to start, and how to tell if your child used AI for homework covers the other side of the same conversation.
If your child's school has not yet told parents how AI fits into revision, this is the work I do with schools: setting a clear AI policy that students and parents can both follow. It is worth sending to your head of year or principal.
I write a Sunday newsletter that keeps teachers and parents up to date without the panic, and parents are very welcome to join it.
Dan Fitzpatrick is a former secondary teacher and assistant headteacher, founder of The AI Educator, and has trained more than 150,000 educators. About Dan.
