Short answer: use AI for finals to make practice exams from your own course material, to explain topics you are stuck on, to mark your attempts after you have answered, and to sketch a study schedule you then make realistic. Don't use it to answer the questions for you during practice. In a field study with nearly a thousand high school maths students, those given an unrestricted chatbot did better while they had it, then 17% worse than students who never had it once it was taken away.
Why the way you use AI matters in finals
Almost everyone is already using it. HEPI's Student Generative AI Survey 2026, based on 1,054 full-time UK undergraduates surveyed in December 2025, found that 95% use AI in at least one way and 94% use generative AI to help with assessed work.
The finals problem is that in the exam hall the AI isn't there. Bastani and colleagues (2025) tested exactly this in a field experiment with nearly a thousand high school maths students. During practice, students with a standard ChatGPT-style tutor scored 48% higher, and students with a version designed to give hints rather than answers scored 127% higher. When access was taken away, the students who had used the standard version did 17% worse than students who never had access. The authors describe students using it as a crutch during practice. The safeguarded tutor largely removed that harm.
The rule that follows: during finals prep, AI should make you produce answers, not hand them over.
Five ways to use AI for finals that actually help
- Make practice exams from your own material. Give it your lecture notes or slides and ask for questions in the format of your final: multiple choice, short answer, problems. In a University of Kentucky guide for finals season, Dr Tama Thé calls practice exams probably the single highest-value use during finals, with the order that matters: answer the questions yourself first, then have AI grade your work and explain your mistakes. Our guide to making a practice test with ChatGPT has the prompts.
- Get a stuck topic explained another way. If the textbook's version isn't landing, ask for a different explanation, an analogy or a worked example. Then close the chat and explain it back in your own words. If you can't, it hasn't landed yet.
- Find your gaps. Paste a list of the topics on your final and rate each one 1 to 3 for confidence. Ask for five questions on each topic you rated 1. The University of Pittsburgh's finals plan suggests the same rating step before you start reviewing, so you begin with the hardest areas.
- Draft a schedule, then fix it. Give it your exam dates and confidence ratings and ask for a day-by-day plan. Then cut it down. As the Kentucky guide puts it, AI schedules tend to assume you're a machine.
- Mark your attempts. After you have written an answer to a past-paper question, ask it to compare your answer with the mark scheme and point out what is missing. Check anything surprising against your notes, because AI marking can be wrong. Our post on whether AI can mark your exam answers explains when to trust it.
What not to use AI for
- Solving practice problems for you. That is the crutch pattern from Bastani and colleagues. If you want help, ask for a hint, not the solution.
- Deciding what is important. The Kentucky guide is blunt that AI is not good for deciding what's important in a course. Your lecturer's guidance, the syllabus and past papers decide that.
- Replacing your notes with its summary. A summary you didn't make is something to read, and reading is the weakest way to revise.
- Facts you haven't checked. Anything that will go into your exam answer should match your course material, not just sound right.
A finals week with AI in it
Pittsburgh's plan starts two to three weeks out with an overview of all the material and a confidence rating for each concept, then moves to self-testing, starting with the difficult areas. With AI, that looks like this:
- Two to three weeks out: list the concepts for each final, rate them, and ask AI to draft a plan. Cut the plan by a third.
- One to two weeks out: each session, a short practice quiz from your notes on the lowest-rated topics, answered without help, then marked.
- The last week: one full practice exam per subject under timed conditions, marked against your notes. Only the misses get reviewed after.
- The night before: a short quiz on your misses, then sleep.
For finals that cover the whole term, our guide to studying for a cumulative final sets out how to divide the material.
Check the rules first
Policies differ between courses and universities. The Kentucky guide's advice is to know what your professor allows and to ask if unsure, with a simple test for the grey area: if you would feel uncomfortable showing your chat history to the professor, that's usually the line. Studying with AI is a different thing from submitting its work, and our guide on using AI without cheating covers where the lines usually sit.
General chatbots can drift from your course. Ace is built for the practice-exam part: it turns your own lecture slides, PDFs or notes into practice questions written from that material only, checks every answer against it and drops questions that fail the check, and shows the sentence from your file under each answer. You pick the chapters, so the quiz matches what your final covers.
Frequently asked questions
Is it OK to use ChatGPT to study for finals?
For studying, usually yes, as long as your course allows it and you use it to test yourself rather than to answer for you. Check your course's AI policy, and ask your lecturer if it isn't clear.
What is the best use of AI for finals?
Practice exams from your own course material, answered without help and then marked. A University of Kentucky finals guide calls it probably the single highest-value use during finals.
Can AI make me worse at exams?
It can if it does the work during practice. In Bastani and colleagues (2025), students who used an unrestricted chatbot on practice problems did 17% worse once access was removed than students who never had it. A version designed to give hints instead largely avoided that.
Should I let AI write my study schedule?
Use it for a first draft, then make it realistic: fewer hours per day, breaks, and the subjects you find hardest earlier in the week. AI plans tend to overestimate what one person can do.
Sources
- Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences, 122(26).
- Stephenson, R., & Armstrong, C. (2026). Student Generative AI Survey 2026 (HEPI Report 199). Higher Education Policy Institute and Kortext.
- Rutland, N. (2026, April 27). Using AI the right way: Tips for finals season. UKNow, University of Kentucky.
- University of Pittsburgh Study Lab. Finals Prep Plan. University of Pittsburgh.



