What the notes were for
The pen-versus-laptop finding wobbled when it was tried again. Two newer studies of AI note-taking suggest what a course should ask students to write for themselves, now that a transcript of every class is free.

Key takeaways
- The 2014 finding that laptop note-taking hurts learning did not hold up in a 2021 direct replication (Urry et al.): no reliable quiz difference between typed and handwritten notes, and the link between copying and worse scores came out shaky.
- In a 30-person University of Michigan study (Chen et al.), ready-made AI notes produced the lowest test scores straight afterwards and were the most preferred. A running summary scored highest, with a live transcript in between.
- In a 405-student trial of 14- and 15-year-olds (Kreijkes et al.), note-taking with or without a chatbot beat the chatbot alone three days later. Most students still preferred the chatbot.
- The rule the edition proposes, untested: let students keep the transcript, since for some it is access, and ask for a short notes page in their own words before any AI summary is opened.
In Maria Cuellar's criminology lecture at the University of Pennsylvania, the screens stay shut for the first half of class and every note goes on paper, coding notes included. The Daily Pennsylvanian reported on 29 October that she is one of a number of Penn faculty cutting technology out of their lectures as AI spreads, and that she says banning screens "helps during lecture" (Dash, The Daily Pennsylvanian, 29 October 2026). The study people usually reach for to back her up has had a bumpy decade.
The pen, the laptop and the replication
In 2014, Mueller and Oppenheimer reported across three studies that students who typed their notes did worse on conceptual questions than students who wrote by hand, and they put it down to laptop users' habit of taking the lecture down word for word instead of reframing it in their own words (Mueller and Oppenheimer, Psychological Science, 2014).
In 2021, Urry and colleagues at Tufts University ran that first study again, as a direct replication, and found no reliable difference in quiz scores between the laptop and longhand groups, in their own data or across a mini meta-analysis of eight similar studies (Urry et al., Psychological Science, 2021). Typed notes were again longer and carried more of the lecture word for word. Longer notes went with better quiz scores, and the link between copying and worse scores, clear in the original, came out shaky the second time. The authors suggest word count and copying may both be stand-ins for a student's motivation or interest (Association for Psychological Science release, 30 April 2021).
So after two well-known studies the pen and the laptop come out about even, and the case against copying is murkier than its reputation. Cuellar may well be right about her room, but these quiz scores won't be what proves it.
Now the record comes free
Whichever way the laptop argument goes, the record itself now turns up without anyone typing it. A student with a phone can get a full transcript of a lecture, and an AI summary of it, without writing a word.
A small study from the University of Michigan tested what happens next. Xinyue Chen, Xu Wang and colleagues recruited 30 people, most of them undergraduates and all of them people who said they found note-taking hard, and had them watch three short recorded lectures, each with a different kind of AI help: a live transcript, a running summary roughly every 15 seconds, or ready-made structured notes every minute or two (Chen et al., CSCW 2025, arXiv preprint 3 September 2025). On a closed-book test straight afterwards, the ready-made notes produced the lowest scores and the running summary the highest, with the transcript in between. The students liked the ready-made notes best anyway, and one admitted to "zoning out." With only the transcript to lean on, 19 of the 30 wrote their notes entirely themselves. It was 30 people tested within the hour, so I wouldn't stretch it past the hour after class.
A larger trial with younger students reaches further. Pia Kreijkes and colleagues at Cambridge University Press & Assessment and Microsoft Research ran a pre-registered randomised experiment with 405 students aged 14 and 15 (Microsoft Research summary; Kreijkes et al., Computers & Education 243, 2026). According to the Cambridge release, the students came from seven secondary schools in England, and each studied two history texts, one with a chatbot and the other by taking notes, either alone or alongside the chatbot, before an unannounced test three days later (Cambridge University Press release, 3 December 2025). Note-taking, with or without the chatbot, beat the chatbot alone on both comprehension and retention. Most students still preferred the chatbot and found it more helpful, according to the Microsoft Research summary. Those were 14- and 15-year-olds working on history texts, so how far it carries to adult learners is an open question.
My hunch, and I'm going further here than any of these papers do, is that the useful part of taking notes is the choosing: sitting through ten minutes of lecture and deciding which two sentences are worth keeping, in words you had to find yourself. A transcript keeps the whole record and leaves the choosing undone. That's fine, as long as something else in the course asks for it. It's a bit like a dashcam, which records every metre of the drive while the driver still has to learn the route by driving it.
A transcript keeps the whole record and leaves the choosing undone.
Where the transcript is the access
One caution before any of this goes into a syllabus. For some students a transcript or a note-taker is an accommodation, and it stays theirs whatever the rest of the class is asked to do. Last year at Concordia University in Montreal, a third-year sociology student, Gabrielle Place, asked the access centre for note-taking support and was offered Genio, an AI service that records classes and produces transcripts and outlines. Her instructors declined consent to recording, and she was later given a human note-taker (Daldalian, The Link, 18 November 2025). For a student whose accommodation is the note-taking itself, any notes page is something to sort out with them and the access office first.
A page of their own
The rule I'd try, in one course, goes like this: the recording and the transcript can be whatever students need, and a short notes page in their own words gets written before anyone opens an AI summary. Cambridge's release says much the same, that students "should take notes separately from using LLMs to avoid copying the LLM."
The transcript is there the minute class ends, and nobody marks it. The notes page gets written before the summary opens, while the student still has to remember what mattered. It fits on an index card, say two ideas worth keeping and one thing still unclear, and three days later, the same delay the Cambridge trial used, it earns one low-stakes question.
How this was made. Full disclosure, and I mean it as a courtesy, not a confession. AI tools help me research, check my claims, tidy my grammar, and stay on top of far more reading than one person can. The thinking and the judgment are mine, I edit every word into my own voice by hand, and then I put my name on it. Telling you that is the whole point of keeping it human.
Here's what I'd love to hear from you, since I can't answer it myself. If your students can get a transcript of every class, what do you still ask them to write down for themselves? Replies from anyone whose students use a transcript as an accommodation would help most, since that's where a rule like this gets tested first.
Next edition: the practice questions students write for themselves, now that an app will happily write them a quiz.
These editions run alongside Learning to Learn With AI, a longer series working the same ground in order.
Common questions
Does handwriting notes beat typing them? Not reliably. A 2021 direct replication of the well-known 2014 study found no clear quiz difference between handwritten and typed notes, and the link between word-for-word copying and worse scores did not hold up the second time.
Do AI-generated notes help students learn? In one small study of 30 people, ready-made AI notes gave the lowest scores on a test straight afterwards, though students liked them best. In a larger trial of 405 secondary students, note-taking beat using a chatbot alone at three days. Both are small and neither tests adult learners over a full course.
Should a course ban lecture transcripts? The edition says no. For some students a transcript or note-taker is an accommodation. The suggested rule leaves the transcript alone and asks for a short notes page in the student's own words before any AI summary is opened.
References
- Dash, A. (2026, 29 October). AI surge prompts more Penn professors to go no-tech. The Daily Pennsylvanian. https://www.thedp.com/article/2026/10/penn-professors-artificial-intelligence-classroom-no-tech
- Mueller, P. A., & Oppenheimer, D. M. (2014). The pen is mightier than the keyboard: Advantages of longhand over laptop note taking. Psychological Science, 25(6), 1159-1168. https://doi.org/10.1177/0956797614524581
- Urry, H. L., Crittle, C. S., Floerke, V. A., Leonard, M. Z., Perry, C. S., et al. (2021). Don't Ditch the Laptop Just Yet: A Direct Replication of Mueller and Oppenheimer's (2014) Study 1 Plus Mini Meta-Analyses Across Similar Studies. Psychological Science, 32(3), 326-339. https://doi.org/10.1177/0956797620965541
- Association for Psychological Science. (2021, 30 April). Don't Ditch the Laptop Just Yet: Replication Finds No Immediate Advantage to Writing Notes by Hand [News release]. https://www.psychologicalscience.org/news/releases/no-benefit-handwriting-notes.html
- Chen, X., Ruan, K., Ju, K. P., Yap, N., & Wang, X. (2025, 3 September). More AI Assistance Reduces Cognitive Engagement: Examining the AI Assistance Dilemma in AI-Supported Note-Taking [Preprint]. arXiv:2509.03392. Accepted at CSCW 2025. https://arxiv.org/abs/2509.03392
- Kreijkes, P., Kewenig, V., Kuvalja, M., Lee, M., Hofman, J. M., Vitello, S., Sellen, A., Rintel, S., Goldstein, D. G., Rothschild, D., Tankelevitch, L., & Oates, T. (2026). Effects of LLM use and note-taking on reading comprehension and memory: A randomised experiment in secondary schools. Computers & Education, 243, 105514. https://doi.org/10.1016/j.compedu.2025.105514
- Microsoft Research. (2025, November). Effects of LLM Use and Note-Taking On Reading Comprehension and Memory: A Randomised Experiment in Secondary Schools [Publication page]. https://www.microsoft.com/en-us/research/publication/effects-of-llm-use-and-note-taking-on-reading-comprehension-and-memory-a-randomised-experiment-in-secondary-schools/
- Cambridge University Press. (2025, 3 December). Note-taking alone or combined with large language models helps students understand and remember better than large language models alone [News release]. EurekAlert! https://www.eurekalert.org/news-releases/1108126
- Daldalian, M. (2025, 18 November). Concordia's AI note-taker draws faculty and student concern. The Link. https://thelinknewspaper.ca/article/concordias-ai-note-taker-draws-faculty-and-student-concern