I praise NotebookLM a lot, and the reason why is simple: it's filled with features that are genuinely useful. Given how impressive all its past updates have been, I've naturally started having high expectations for each new release. Unfortunately, the Deep Research feature Google recently added to NotebookLM didn’t live up to those expectations.

What is NotebookLM's Deep Research?

Gemini and other AI tools had it first

If you've used any of the major AI tools out there, like Perplexity, ChatGPT, and even Google’s own Gemini, you might’ve noticed that they all have dedicated Deep Research modes. Deep Research is designed to do exactly what you’d expect — dig deeper into a topic. While the default research mode typically only gives you quick summaries or surface-level answers, Deep Research is built to go several layers further.

It searches across tens to hundreds of websites, including community forums like Reddit, to gather different perspectives and give you a more complete picture of whatever you're researching. Before Deep Research actually begins, it even presents an in-depth plan that you can edit or approve. And once it’s done, it generates a full report summarizing its findings, breaking everything down into neat sections so you can review it at a glance.

On the 13th of November, Google expanded the feature beyond its initial scope and added Deep Research directly to NotebookLM. It has two research style options: Fast Research and Deep Research. The former option is meant for quick searchers and to rapidly find you new sources, while the latter is the full-fledged Deep Research mode.

Users on Reddit had already spotted the feature in testing, and frankly, I was all in. I frequently save the Deep Research reports Gemini generates and add them to NotebookLM to take advantage of its better Audio Overviews and other great learning features. So, naturally, I was excited to see how NotebookLM would handle Deep Research natively.

NotebookLM’s Deep Research feels like it's drifting from the app’s core identity

It doesn’t quite fit the app’s core idea

Since NotebookLM's early days when it was being tested in Google Labs, the reason why the tool was so well received was simple: it stayed focused on your sources. Everything it generated had to be grounded in whatever you uploaded. Since this considerably reduced hallucinations, it was what made NotebookLM feel dependable in a way most AI tools don’t.

Though NotebookLM has gotten countless major updates since then, its source-grounded nature has stayed mostly intact. While the original Discover Sources feature (which is where you’ll now find Deep Research) did search the web to pull in additional material, it was shoved into the background, never feeling like the main focus. It supplemented your uploads rather than taking center stage, keeping the app’s source-first identity intact.

Instead of the Discover Sources button being tucked away in a spot that was easy to ignore, Deep Research now takes a much more prominent role in the interface. It’s front and center in the Sources panel. Of course, you can choose to ignore it and stick to just your uploaded documents, but the placement makes it hard not to notice.

Additionally, the key to using NotebookLM efficiently is to populate your notebooks with high-quality sources. The issue with Deep Research is that it isn't always the best at determining which sources are truly relevant or credible. I've had many Deep Research reports cite Wikipedia, outdated articles, or sources that barely touch on the topic I'm exploring. While these can sometimes provide useful context, they often require extra vetting before they’re actually usable in a notebook.

The addition of Deep Research also makes me worry about NotebookLM's identity as a whole. By pushing external reasoning (which is what Deep Research is) to the forefront, NotebookLM could very well be drifting away from the very quality that sets it apart. That said, I do hope I’m wrong about that.

Deep Research doesn’t work on your Drive yet

The feature that would've justified Deep Research’s existence

As I mentioned above, I'm not the biggest fan of Deep Research being a feature designed to fetch new sources — which is the opposite of NotebookLM's biggest strength. Early this month, Google announced that Gemini's Deep Research can draw context from Gmail and Google Drive. An example Google mentioned in its The Keyword blog perfectly sums up the ideal use case:

Now you can start a market analysis for a new product by having Deep Research analyze your team's initial brainstorming docs, related email threads and project plans. Or you can build a competitor report about a rival product that cross-references public web data with your strategies, comparison spreadsheets and team chats.

Now, given that NotebookLM is specifically designed to work with your sources, this feels like the exact kind of upgrade Deep Research should have launched with. If it had, I wouldn't have been nearly as skeptical about its place in the tool. In fact, the ability to run deep, context-aware research on Drive files would've only been a natural extension of what NotebookLM already excels at. It would’ve strengthened its core identity instead of pulling the experience in a different direction.

NotebookLM's Deep Research feels half-baked

Shipped before it was truly ready

When you're using Deep Research in tools like ChatGPT, Perplexity, or Gemini, your goal is typically to get a comprehensive answer without spending hours digging through sources yourself. With Deep Research in NotebookLM, the purpose is slightly different. In addition to using the report it generates, you also want to integrate the insights you find directly into your notebook.

The Deep Research feature is designed to help you find new, high-quality sources for your notebook that align with the topics you're exploring. Given that the focus is on finding sources, I'd expect that using the sources Deep Research finds would be easy. Unfortunately, that doesn't seem to be the case. When Deep Research is done conducting its research and generating the report, all the sources it cited in the report (and some it didn’t) are listed at the bottom with a checkbox next to them. By default, they're all checked and ready to be added to your notebook.

However, with NotebookLM, the sources you add to your notebook are extremely important. So, whenever I’m adding sources to my notebook from external places, I make sure I vet them carefully to ensure they’re credible and relevant. With Deep Research, though, the process feels clunky since you need to open each source it finds in a separate tab. You can't view any of them within the interface itself, which means you have to constantly switch between tabs just to see if a source is actually useful.

When you generate a Deep Research report using Gemini, there's an Export to Docs button that lets you save the entire report in a Google Docs format. This makes it extremely easy to edit the report, share it with others, and upload it to NotebookLM notebooks.

Unfortunately, in the NotebookLM version of Deep Research, the report is displayed as a markdown (.md) file. Weirdly, you can't download the markdown file! You can view the report as-is within NotebookLM, yet you can't currently export it.

Sure, you can copy and paste it into a Google Doc yourself, but it's just an unnecessary step that shouldn't be there. Doing so also seems to mess up formatting, which means you need to go in and clean it up yourself. The lack of an export option means you can’t really reuse reports as sources in other notebooks without manual work. While these are minor annoyances, they add up and create unnecessary friction in your workflow.

I still have hope the feature will improve

Given NotebookLM's past record (and the impressive features it launched recently, like Slides and Infographics), I still have confidence Google will improve the feature. While it does feel half-baked right now, the potential is there for it to truly complement NotebookLM's core strengths.

However, my main point still stands nonetheless: as it exists today, Deep Research feels like a feature that pulls NotebookLM away from what originally made it unique. Until it better integrates with your own sources, the tool’s core identity as a source-grounded research assistant remains somewhat compromised.