NotebookLM is one of the few generative AI tools that has really impressed me, but this doesn't mean that it doesn't have flaws. There's a lot it does better than its competitors, including letting you set the sources you want to use.
That said, it also has drawbacks, especially with some of its newer features. Here are a few things I think NotebookLM should work on to improve the tool's usability.
The bare-bones source view
Revisiting sources is not exactly seamless
When Google first announced its partnership with OpenStax to bring some of its most popular textbooks to NotebookLM, I was really excited to jump into the content. I love learning new things, especially when it comes to interests I wasn't able to pursue during university. One of these is psychology, so I was delighted to see that a notebook introducing psychology was included in the partnership.
However, this is also when I noticed how limiting NotebookLM's view of sources is. I didn't really want to ask the chatbot questions about the content until I had read it myself. But viewing sources within NotebookLM is really inefficient.
Firstly, the Sources pane can only take up about a third of the screen. If you close the Studio pane, it only expands the footprint of the Chat pane. You can zoom in on your browser until the chat disappears, but by the time this happens, you're not seeing much text from the sources.
Then there's the formatting. Sources are mostly formatted in plain text, so you lack the benefits of header formatting. You get the occasional bold and italic text, but it actually doesn't stand out much from the rest of the text. The OpenStax chapters also included occasional images, but far fewer than the original documents. I eventually decided to just use the original OpenStax resources rather than NotebookLM.
But my experience made me notice just how limited the source view is. When you upload documents from the web or your Google Drive, you can at least visit the original link to see the source in its original formatting, along with images. But if you upload a PDF directly to NotebookLM, the scraped text is the only copy you have.
The quality of the text scraping also seems to differ depending on whether you use direct upload or Google Drive, even when uploading the same document. I think this really hampers the potential of NotebookLM as a learning resource for material you're completely unfamiliar with, if you would like to first view the original information.
Revising videos and audio according to prompts
Not much changes with certain prompts
One of the ways I've used NotebookLM is to explain my complex chronic illnesses to family members in an accessible way. I created a video explaining chronic migraines and their causes to help them better understand the challenges of treating the condition.
The first video, however, was very general. I hadn't entered a prompt, so I decided to create a new video with some specific prompts. This included asking the chatbot to explain what treatments were used for people who had failed first-line and second-line preventative treatments, as this is the situation I'm dealing with.
However, the newly generated video barely addressed this point, devoting less than 30 seconds to it in the nearly 8-minute video. It mostly just reiterated the same points as the first video. I also checked that the chatbot could generate this information in the chat, which it could, meaning it wasn't due to a limitation in the sources.
In fact, the biggest difference between the videos was that the narrator changed from a masculine to a feminine voice. I decided to go with the second video just because the narrator was less loud, which helps with my noise sensitivity.
I later tried to generate a video focusing specifically on the link between chronic migraine and fibromyalgia. Once again, I confirmed the chatbot was able to pull this information from the sources through a text prompt. However, when I generated a video, the video once again covered the same ground. It defined the difference between chronic and episodic migraine, and used the same line about migraine not being "just a really bad headache".
Almost halfway through the video, it eventually got to the topic of comorbidities. But then it wasted time explaining the definition of comorbidities. After three minutes, it listed the common comorbidities. Over half the video was done before it started, discussing what I had prompted: the link between chronic migraine and fibromyalgia. It briefly discussed this, then recycled information from the other videos.
Video overviews are still an impressive tool, but if they really want to be useful, they need to truly tweak their output to prompts, rather than recycling the same content with minimal changes.
YouTube video citations
They're lacking essential features
NotebookLM does a great job of citing its answers. It also makes it easy to locate the relevant parts of documents for citations, especially text sources, by highlighting the cited information. However, for video sources, the chatbot is much less effective.
When you click a citation from a video source, NotebookLM opens the embedded video and its transcript. However, it doesn't provide timestamps, which makes it difficult to find the relevant part of the video. I have also noticed it highlighting the entire transcript multiple times, rather than the most relevant part. In these cases, hovering over the citation also caused the pop-up to display the entire transcript rather than a focused excerpt. I've encountered this issue across multiple notebooks, so it's not just limited to one or two cases.
While I mostly use text sources with NotebookLM, I first noticed this shortcoming when creating a customized workout routine with the chatbot. It was also frustrating when using the chatbot for certain gaming guides. I imagine that if you're dealing with particularly long videos, the lack of timestamps can become especially frustrating.
The ineffective Android app
It's disappointing compared to the web version
NotebookLM is a cloud-based app, so I'd argue that mobile functionality is important. However, the mobile experience, at least what I've tried on Android, pales in comparison to the web version. Now it's worth noting that the store page notes that this is an early version of the NotebookLM app, but it has been out since May 2025, so I expected that it would be further along than it is.
My main gripe with the app is that while it syncs the notebooks you have saved, it doesn't sync most of your saved Studio output. I can find the occasional audio overview, but my mind maps, FAQs, and saved responses are not available in the app. I understand it not having access to some of the latest features like quizzes and flashcards, but it's frustrating when mind maps and notes aren't saved either.
In addition, featured notebooks I've accessed out of curiosity are mixed in with my actual notebooks, with no way to remove them. Meanwhile, the source view is even more limiting because there is no way to visit the original website you have referenced.
The limitations of NotebookLM are essentially amplified when it comes to the mobile app.
Obtrusive featured notebooks
They feel a bit like ads
Featured notebooks were rolled out to NotebookLM in mid-2025. While I was initially excited to see what they could offer, I was left underwhelmed by the actual experience. But eventually, these notebooks were more than just underwhelming; they were actually annoying.
This is because Google has placed them at the top of the NotebookLM home page, above your actual notebooks. This means that the first thing you're greeted by is featured notebooks, which have not seen any additions since July. This applies both to the grid and list view, as well as if you sort your notebooks by most recent.
As a result, notebooks I have never visited get precedence on my home page over my own saved notebooks. Since they're rarely updated, I can't even say that this is a feature that helps you discover new content. Rather, it just feels like a way to advertise books and publications.
Newer partnerships, like the one Google has with OpenStax, don't even feature in this section. And in the mobile app, if you've visited one of these notebooks, they remain on your home page with no way to remove them.
I would honestly prefer that this section be replaced by pinned notebooks, notes, and overviews, rather than something that doesn't feel useful at all.
NotebookLM still has room to improve
There's a lot I like about NotebookLM, and I'll continue to use the tool to summarize large amounts of information in a digestible way. At the same time, I hope that Google fixes these issues, since it is one of the most promising experiments to come out of Google Labs.