Why does writing while listening feel harder than either one alone?
This comes down to a basic constraint of working memory. Cognitive Load Theory treats working memory as a limited resource with three competing demands: intrinsic load (how inherently complex what you're hearing is), extraneous load (unnecessary difficulty added by how the task is structured — like having to write and listen at once), and germane load (the effortful processing that actually builds understanding).
When you're taking notes during a live conversation, you're not just listening — you're also deciding what's worth capturing, translating it into your own words or shorthand, and physically writing or typing it, all while the conversation keeps moving.
Direct experimental work on divided attention backs this up outside the classroom context too: dividing attention between two simultaneous streams of information — even something as basic as two spoken sentences — measurably increases processing effort and reduces accuracy on both, compared to attending to just one.
Does the specific method of note-taking — typing vs. handwriting — change the picture?
It shifts the trade-off rather than eliminating it. Research on note-taking modality has found that when people write by hand, one specific factor — the speed at which someone can physically write — has been a significant predictor of both note quality and later recall.
Typing is faster, which sounds like an advantage, but it comes with a documented downside: people who type tend to default toward capturing information verbatim rather than actually processing and reformulating it in their own words, and research connects that verbatim-transcription tendency to shallower processing of the material, not deeper engagement with it.
The honest takeaway from this line of research: there isn't a note-taking method that removes the underlying cost. There's only a choice about which cost you're willing to accept — slower, more effortful capture that processes information more deeply, or faster capture that risks becoming closer to passive transcription.
Does simply automating note-taking with AI solve the problem?
This is where the research gets genuinely interesting, and where the obvious assumption — "removing the physical writing removes the cost" — turns out to be only partly right. A rigorous 2025 study published through the ACM specifically tested this by giving 30 participants three different levels of AI note-taking assistance while they followed a live lecture: fully automated structured notes, an intermediate level offering real-time summaries they had to actively select and integrate, and a minimal level offering only a raw transcript.
The result cuts against the obvious assumption: participants using the most automated, hands-off AI assistance scored the lowest on a comprehension test afterward — significantly lower than the intermediate condition, even though they rated the fully automated notes as higher quality and easier to use. The intermediate condition, which still required people to actively read, select, and integrate AI-generated content rather than just receiving a finished product, produced the best comprehension of the three.
The researchers describe this as a mismatch between what people prefer (maximum convenience) and what actually produces the cognitive benefit they're trying to get from taking notes in the first place.
So what does the evidence actually recommend?
Put together, this body of research points to something more specific than either "always take your own notes" or "let AI handle everything":
- > Removing the physical act of writing does free up attention for the actual conversation — that part of the "let AI transcribe it" instinct is well-supported. Divided-attention and cognitive-load research is consistent that competing for the same working-memory resources hurts both tasks.
- > But removing your active engagement with the content entirely appears to cost you the comprehension benefit that note-taking was providing in the first place. The 2025 ACM study's finding — that full automation produced the worst understanding of the material, not the best — is a direct, specific data point against the assumption that more AI assistance is strictly better.
- > The sweet spot the research points to is being handed digestible pieces to actively select from, not a finished product to passively receive. That's a meaningfully different design goal than "generate a perfect summary automatically" — it's closer to "surface the right raw material and let the person do the light, active work of choosing what matters."
Our companion post on what a record-first workflow actually looks like goes into what that means in a real client meeting, once you're not the one holding the pen.
Key takeaways
- > Note-taking and listening draw on the same limited working memory, so doing both at once measurably degrades both.
- > Handwriting processes material more deeply than typing, but no note-taking method removes the underlying attention cost.
- > In a 2025 ACM study, the most automated AI note-taking produced the worst comprehension of the three conditions tested.
- > The best results came from actively selecting and integrating AI-organised material, not from passively accepting a finished summary.
Sources
- Chen, X., Ruan, K., Ju, K. P., Yap, N., & Wang, X. (2025). More AI Assistance Reduces Cognitive Engagement: Examining the AI Assistance Dilemma in AI-Supported Note-Taking. Proceedings of the ACM on Human-Computer Interaction, 9(CSCW451). arxiv.org/pdf/2509.03392
- Klepsch, M., Schmitz, F., & Seufert, T. (2017). Development and validation of two instruments measuring intrinsic, extraneous, and germane cognitive load. Frontiers in Psychology, 8.
- Kiewra, K. A. (1989). A review of note-taking: The encoding-storage paradigm and beyond. Educational Psychology Review, 1(2), 147–172.