As a regular NotebookLM user, here are 5 things I want from Gemini Notebook

A few weeks ago, Google announced it was rebranding NotebookLM to Gemini Notebook. Initially, this change was purely cosmetic, but we’re already seeing signs that Google has big plans for the future.

In fact, we’ve already seen a few new features like Collections, which finally lets you group related notebooks together like a playlist. That’s just the beginning, though, as a recent APK breakdown hints the tool may soon let users add website links as sources directly in the Gemini app, alongside a few other possible changes.

This got me thinking about what else I’d like to see from Gemini Notebook in the future. Here are just five relatively modest yet important changes I would personally love to see introduced into the AI tool.

Multi-notebook merging

While one notebook is often enough, certain projects can benefit from multiple workspaces, making it easier to keep things organized during the early planning stages. For example, a graduate student might be working on a thesis with a ton of sources and information to sift through. To make things easier, they might break things down into separate notebooks for each chapter. This is useful early on, but less useful once you know exactly what you’re doing and where you are heading. Imagine if there were an easy way to merge these related notebooks together.

Unfortunately, there is no official way to merge multiple notebooks directly in the Gemini Notebook app. There are workarounds, such as third-party tools. You can even upload or reference multiple notebook links in Google Gemini and then ask it to cross-analyze, synthesize, or combine content across all attached notebooks. All of these methods require additional steps and the use of tools outside the Gemini Notebook workspace. A native merge feature would be a natural fit for the AI-powered tool, but there are some reasons it hasn’t happened yet.

The main reason this feature doesn’t exist yet is that Gemini Notebook is designed around strict source grounding to prevent AI hallucinations. Mixing several independent notebooks greatly increases the odds of hallucinations and other issues. That said, this should become much more achievable in the future as Google continues to improve the underlying AI technology and further expands notebook source caps.

The ability to download sources

Imagine spending tons of time putting together a research project in Gemini Notebook, but down the road, you find you need to reference your sources again. While you should ideally already have copies of these somewhere, that doesn’t mean you always will.

Right now, there’s no easy way to download the sources you previously uploaded, such as PDFs and other documents. Just as with notebook merging, there are workarounds, but none are perfect. Probably the best method is to use Google Takeout to request an archive of your data, which makes it possible to get all your old sources manually. Still, a one-click export for individual sources would be extremely easy to add and a lifesaver on the rare occasion where you lose your original source material.

More ways to learn or recap information

I love how Gemini Notebook makes it easy to generate interactive quizzes, visual diagrams, slide presentations, and even podcasts around your notebook sources. That said, I think there’s room to expand beyond this.

Imagine if it could take your sources and turn them into more deeply interactive resources, like an interactive quiz that lets you compete against others—more like a game show than a study-focused version. I’d also like to see short games and such, too. Despite the screenshot, I’m not talking about a Minecraft level. My vision here is more like simple learning games from the 90s, or even the classic Oregon Trail.

To be fair, this is probably a bit of a bigger ask than game show-style quizzes, at least at the depth I’m describing here. The good news is that the bones of this idea might already be forming.

Recently, Testing Catalog spotted a new menu option called the “App within the Studio” section. This seems to expand on existing options like Infographics, Mind Maps, and Flash Cards, as it mentions the ability to create “an interactive web page,” “a fun game,” or a “visual dashboard or interactive tool.” It’s unclear how deep these “fun games” will be. They could be simple text-based affairs, or something a bit more. Still, it’s a start.

Direct database connectors

Imagine if Gemini Notebook integrated live-sync connectors for tools like Obsidian, Slack, or GitHub repositories, rather than static, manual file uploads. This means notebooks would automatically update as the live operational databases are modified. It would also make it easier to ask questions that combine structured text with unstructured data, like inventory metrics or customer transactions. Supporting direct database connections like this would also mean users could bypass the existing, relatively strict file-volume limits.

As much as this would transform the notebook and expand its capabilities, it would also be hard to implement within the tool’s existing retrieval-augmented generation architecture, as it’s not well designed to handle aggregated data. Tying to live databases could also cause potential security vulnerabilities. As much as I’d love to see this one, I mostly get why Google hasn’t gone this route.

Expanded PDF size support

Gemini Notebook supports up to 500,000 words per individual source, or about 1,000 pages single-spaced. There’s also a 200MB file limit per upload. That sounds great, and for basic users like me, it’s more than enough. However, that might be less true for bigger businesses, academic research, and other high-level use cases.

Besides, even if you can technically upload large documents, the reality is that Gemini Notebook can’t always fully parse them. If you were to search for a specific topic, the notebook would simply create snippets based on the search parameters, but certain terms could max out before it actually reads the whole document. This could result in missing information.

The good news is that there are ways around this, such as breaking bigger documents into smaller, readable chunks. You can also use markdown, reinforce parent-child relationships with headings, and other methods to make it easier to scan what you really need. This can improve the analysis quality, but it’s also a lot more work.

I get that AI computing costs are probably the main limit here, but I do think raising the limits would make sense for those with a higher-tier subscription, such as business owners.

Thankfully, Gemini Notebook is still mostly great as is

As much as some of the above would be nice, none of them are absolutely necessary or deal breakers for me. The truth is that there might be a lot of somewhat similar tools out there, but the full Gemini Notebook experience is hard to fully replicate. I’ll be honest: if I had made this list a year ago, it would have been much larger. Google has done a great job improving the tool, and I’m sure it will only get better under its new branding as Gemini Notebook.

Are there any features I didn’t mention that you’d love to see? We’d love to hear about them down in the comments below.