Study tips12 min read

How to use AI to revise at university: lecture slides, readings and your university's AI rules

Explore prompts that quiz you, UK university AI policies, how AI use is checked and the Feynman technique with a chatbot.

Medly Team19 November 2026
On this page

Yes, you can use AI to revise at university, and UK universities generally allow it for revision even when an assessment bans it. The method that works best is to have AI quiz you on your own lecture slides and set reading, then check its answers against your module materials. This guide gives you the prompts, the checks, and what your university's AI rules mean for revision and coursework.

Can you use AI for revision at university?

Yes. In July 2023 the 24 Russell Group universities agreed five principles on generative AI, and the first is to help students become AI-literate. The same document says appropriate uses "are likely to differ between academic disciplines", so the detail is set by your university, your department and each assessment.

Most universities sort assessments into a few categories. UCL uses three categories. In Category 1 you can't use AI in the assessment itself. In Category 2 it can play an assistive role, such as proofreading. In Category 3 it is part of the task. Even for Category 1, UCL's student guidance says you might use AI to help you revise. Leeds uses a similar red, amber and green scheme.

Revision and coursework sit under different rules because revision produces nothing you hand in. Coursework does, and the assessment brief decides what counts as your own work.

How you use AIRevisionCoursework
Quizzing you on your notesUsually fineFine as preparation
Explaining a concept againUsually fineFine as preparation
Proofreading a draftNot relevantDepends on the brief
Writing text you submitNot relevantMisconduct unless the brief allows it

Source: Russell Group, UCL and University of Sheffield guidance. Your assessment brief always takes priority. Last checked: 18 November 2026.

Before you upload anything, check what you're allowed to share. Lecture slides usually belong to the university, and your notes may include personal details from seminars or placements. Sheffield's revision guidance points students to Google's NotebookLM and asks them to work through its "Should I upload this to AI?" checklist first. Start with the tools your own university supports, and check the data settings of any other tool before you use it.

Which AI tools can you use to revise at university?

Most students start with a general AI chatbot. It can explain almost anything, but it wasn't built for studying. You write the prompts every session, and keeping track of what you got wrong is up to you.

Medly is built for revision. You upload your lecture notes, slides, readings and essays for each module, and Medly does the setting up:

  • It writes a summary of each lecture that you can edit, so you can correct anything that doesn't match how your lecturer taught it.
  • It builds personalised flashcard decks from your own material.
  • It has a large question bank and practice papers, with practice tests built from your lectures in the question style your exams use.
  • It marks your answers and gives you personalised feedback on what you missed.
  • Its chat lets you ask about any concept or highlight part of a slide, with your course materials as context.

For coursework, Medly helps you plan an essay using your own materials, but it won't write the essay for you. If your brief allows AI feedback, you can upload a draft to get it marked.

ToolBuilt forKey features
General AI chatbotAnswering any questionAnswers and explanations from the prompts you write
University-supported toolVaries by universityCheck your university's guidance on what to upload
MedlyRevision and examsUploads, editable summaries, personalised flashcards, question bank, practice papers, personalised marking, chat

Source: Medly for University and University of Sheffield guidance. Last checked: 19 November 2026.

How to use AI for university revision: turn each week's lectures and reading into questions

Reading an AI summary is still rereading, just with fewer words. Testing yourself works better. In a 2006 study by Roediger and Karpicke, students who practised recalling a passage remembered more of it a week later than students who spent the time rereading it.

So ask AI for questions instead of summaries. Once a week, for each module:

  1. Gather that week's slides, your own notes and the set reading.
  2. Upload them, or paste in the sections you need, with the prompt below.
  3. Answer every question from memory before you look anything up.
  4. Note the questions you got wrong and ask them again three days later.

I'm revising week 6 of my [module name] module. Using only the slides and reading I've uploaded, ask me 8 questions, one at a time. Mix recall questions, "explain why" questions and questions that apply an idea to a new case. Wait for my answer, then mark it, tell me what I missed and say which slide or page the answer comes from.

The last line matters. Asking for a slide or page number ties each answer to your course and gives you something to check. Once the weekly quiz is a habit, add the other four prompts:

PromptWhat to ask forA good answer
Weekly quiz8 questions from your slides, one at a timeCites a slide or page each time
Explain why5 "why" questions, each with a follow-upPushes you past definitions
New caseA scenario you haven't seen beforeMarks your answer against your notes
Gap checkThe gaps in your own explanationAsks questions instead of rewriting
Source checkThe exact sentence behind an answerSays so when it can't find one

You can download the prompt sheet with all five written out to copy and paste. If you'd rather not write prompts every week, upload your lectures, slides and readings to Medly. It writes an editable summary of the week, builds a flashcard deck from it and gives you practice questions to answer and get marked.

Checking AI output against your module handbook, slides and set reading

AI tools can sound sure of themselves and still be wrong. Sometimes they're also right about the subject in general but wrong for your course. Sheffield's guidance tells students to "always review the output received critically", and the checking counts as revision too.

For each answer, check three things:

  • Does it match the slide or page it cites? If that reference doesn't exist, treat the whole answer with care.
  • Does it use your lecturer's terms and models? A textbook definition can differ from the one your module teaches, and your exam is marked against the module.
  • Does it fit the learning outcomes in your module handbook? They tell you what the exam can ask, so an answer that goes beyond them is less useful.

Watch for answers from the wrong country too, such as US case law in an English law module. If you find an error, correct it in your notes in your own words. That one fix will stay with you longer than ten correct answers you skimmed. In Medly, you can make the correction in the lecture summary itself, because summaries are editable.

How do universities check for AI use?

Many universities run coursework through Turnitin, whose AI writing indicator estimates how much of a submission may be AI-generated. Turnitin's help centre says it shows an asterisk instead of a number when that estimate is between 1% and 19%, because low scores are less reliable.

Universities say the score alone isn't proof. Bristol's guidance says it "should not be interpreted as proof that AI has been used" and must be weighed alongside academic judgement and the wider context of the assessment. If a concern is raised, a university can look at other evidence, such as your drafts, your references and whether you can talk through your work.

None of this affects revision. Quizzing yourself with AI leaves nothing in your submitted work to detect. The thing to get right is that anything you hand in follows its brief.

What is the 30% rule in AI?

The "30% rule" is an online claim, not a university rule. It's usually explained in one of two ways: that up to 30% of an essay can come from AI, or that a detector score under 30% is safe. We haven't found a UK university policy that uses either figure. The rule that applies to you is the one in your assessment brief, and in an assessment where AI isn't allowed, any AI-written text breaks it.

Feynman with AI: explain it yourself, then find the gaps

The Feynman technique has two halves. First, explain a concept in plain words, as if to someone new to the subject. Then find the places where your explanation goes vague, because those are your gaps. AI is good at the second half.

  1. Pick one concept from this week's lectures.
  2. Without notes, write an explanation of 100 to 150 words.
  3. Paste it in with the prompt below.
  4. Fix the gaps in your own words, then explain it again the next day.

Here is my explanation of [concept] for my [module] module. Don't rewrite it. List the gaps: anything missing, vague or wrong. For each gap, ask me a question that would help me fix it myself.

"Don't rewrite it" keeps the thinking with you. You can use the same prompt in Medly's chat, where it can draw on the lectures you've uploaded for that module.

Worked example

A first-year psychology student explains the working memory model:

"Working memory is short-term memory. Baddeley and Hitch said it has a central executive that controls the phonological loop and the visuospatial sketchpad. The episodic buffer was added later."

A good response lists gaps like these:

  1. "Working memory is short-term memory" blurs the main point. The 1974 model replaced a single short-term store with several components that process information as well as hold it. Question: what does working memory do that a simple store doesn't?
  2. It doesn't say what each component holds. Question: what kind of information goes to the phonological loop, and what goes to the visuospatial sketchpad?
  3. There's no evidence. Question: what do dual-task studies show about these two components?
  4. "Added later" is vague. Question: when was the episodic buffer added, and what does it do?

Compare the gaps with what your lecturer stressed, and you know where your next session should go. For the wider picture of a first-year course, see our guide to psychology's main approaches.

Working memory model practice questions with answers: a sample AI quiz for psychology

These are the kind of questions a good prompt should produce from a week of lectures on memory. Try each one before you open the answer, then compare the mix with what your own AI quizzes give you.

Question 1: (1 mark)

Which component of Baddeley and Hitch's working memory model holds speech-based information?

A. Central executive

B. Phonological loop

C. Visuospatial sketchpad

D. Episodic buffer

Answer

B. Phonological loop

It holds sound and speech-based information. The visuospatial sketchpad holds visual and spatial information.

Question 2: (2 marks)

Name the component Baddeley added to the working memory model in 2000, and state its function.

Answer
  1. The episodic buffer (1 mark).
  2. A limited-capacity store that combines information from the other components and long-term memory into single episodes (1 mark).

Question 3: (4 marks)

People can usually track a moving light on a screen while saying "the the the" aloud. They remember fewer words from a list they've read if they say "the the the" while learning it. Use the working memory model to explain both findings.

Answer
  1. Tracking the light uses the visuospatial sketchpad (1 mark).
  2. Saying "the" aloud uses the phonological loop (1 mark).
  3. The two tasks use separate components, so they interfere very little and both can be done together (1 mark).
  4. Remembering words relies on rehearsing them in the phonological loop. Saying "the" blocks that rehearsal (articulatory suppression), so fewer words are recalled (1 mark).

Marks are usually lost by saying one task is "harder" without naming the component each task uses.

Question 4: (4 marks)

After a motorcycle accident, patient KF had poor short-term recall of spoken letters and numbers. His recall of visually presented information was much better, and his long-term memory was largely intact. Explain how this case supports the working memory model, and give one limitation of using it as evidence.

Answer
  1. The damage affected short-term memory for verbal sounds but not for visual information, which suggests these are separate components (1 mark).
  2. This fits a damaged phonological loop alongside a working visuospatial sketchpad (1 mark).
  3. Limitation: KF is a single case, so the findings may not apply to people in general (1 mark).
  4. There's no record of his memory before the accident, so we can't be sure the difference was caused by the damage (1 mark).

Question 5: (2 marks)

A chatbot told a student: "The central executive is the main store in working memory and holds information from the other components." Identify the error and correct it.

Answer
  1. The central executive isn't a store (1 mark).
  2. It directs attention and coordinates the other components, and has very limited capacity of its own (1 mark).

This is the kind of answer to check against your slides before you learn it.

To practise more questions like these on your own modules, upload your lectures, slides and readings to Medly. It builds practice questions and flashcards from them, and marks your answers so you can see what to go back to.

Frequently asked questions

How do you use AI for university revision?

Ask AI to quiz you on your own lecture slides and set reading instead of summarising them. Answer from memory, then check each answer against the slide or page it cites and your module handbook. Only use the tools and uploads your university's guidance allows.

Can you use AI for revision at university?

Yes. UK universities generally allow AI for revision, even for exams where it's banned in the assessment itself. Check your university's guidance for the tools it supports and what you can upload.

How do universities check if you use AI?

Many run coursework through Turnitin's AI writing indicator. Universities such as Bristol say the score isn't proof on its own and weigh it with academic judgement and other evidence.

What is the 30% rule in AI?

It's an online claim that up to 30% AI content is acceptable. We haven't found a UK university policy that uses it. The rule that applies is the one in your assessment brief.

Is it cheating to use AI to make flashcards from my lecture notes?

No. Flashcards for your own revision aren't submitted work. Writing some of them yourself is still worthwhile, as Sheffield's guidance points out, because making them is revision too.

Can I upload my lecture slides to an AI chatbot?

Check first. Slides usually belong to the university and some notes contain personal details, so use the tools your university supports and follow its upload guidance.

Can Medly help with university revision?

Yes. You upload your lecture notes, slides, readings and essays, and Medly gives you editable summaries, personalised flashcard decks, practice questions and papers, personalised marking and a chat to ask about any concept.

Try Medly for free