NotebookLM has come a very long way from where it started. It began with a simple Chat window where you could throw a few sources in and ask questions, and answers would come back with citations pointing right at the line you needed. That alone was enough to make it stick, since it actually made AI useful to the average user.

Then came Audio Overviews, and NotebookLM went absolutely viral. The team has been constantly shipping features left and right, and improving existing ones too. Mind Maps is one feature that's stayed practically the same since it launched — and that finally changed a few days ago. Turns out, an update was all this feature needed to go from useless to something I can't survive a study session without.

NotebookLM's Mind Maps are finally customizable

It only took Google a year, huh

As someone who has been using NotebookLM since its early Google Labs days and follows its development closely, I've noticed a pattern in the way the company rolls out features. Every new Studio feature almost always begins with a simple tile in the Studio output panel, and all you can do is click it and the tool spits out an output. Features typically don't ship with customization options right out of the box, and the team tends to wait, watch how people actually receive the feature, and then add improvements. Audio Overviews launched as a one-click button that gave you a podcast about your sources, and only later picked up length controls and language options.

Almost every other NotebookLM feature followed the same pattern, and I assumed it'd be the same with Mind Maps. Mind Maps is coincidentally the first NotebookLM feature I covered as news for XDA, on the 19th of March 2025, and it only just got its first meaningful upgrade now, well over a year later. On May 5th, NotebookLM announced via a post on its X account that you can now customize your Mind Maps with specific user prompts. In practice, Google implemented it the same way all the other Studio outputs are customizable. There's now a small pencil icon next to the Mind Map tile, and clicking it opens a text panel that asks What should the topic be? You type your prompt in, hit generate, and the map that comes back is shaped by what you asked for instead of whatever NotebookLM thought your sources were broadly about.

If you'd like a Mind Map that covers everything, you can do exactly what you've been doing all along and just hit the generate button without typing anything. If the old auto-generated map worked for you, you can keep generating those exact same maps forever. But if you've ever looked at one and thought this isn't quite what I needed, you now have a way to tell NotebookLM what you actually needed.

Mind Maps can finally find the connections that actually matter

Cross-source synthesis, minus the noise

The actual reason I kept opening Mind Maps in the first place, even when the output was underwhelming, was the cross-source synthesis. That's the one thing a Mind Map can do that none of NotebookLM's other Studio outputs can't — show you how ideas from different sources connect to each other. A Mind Map does just that visually in a way you can actually trace.

The problem was that the old Mind Maps tried to find every connection at once. If you had three sources in a notebook, the map was usually fine. If you had fifteen, it became a sprawling web of every possible relationship across every possible topic, which is functionally the same as showing you nothing. The connections were technically all there. You just couldn't see the ones that mattered for whatever you were actually working on.

Prompting fixes this almost entirely. Instead of asking NotebookLM to map everything, you can now point it at the specific thread you're trying to follow. The Mind Map that's produced with a prompt is much more narrow, focused, and actually readable. The connections it surfaces are the ones I came looking for, rather than the ones the algorithm thought were representative of what a user would want. For instance, if I have a notebook with a bunch of lecture slides and I want to see how a specific concept shows up across different units, I can prompt the Mind Map to focus only on that.

I still wish you could edit Mind Maps

For all the credit this update deserves, there's one limitation that hasn't budged: the maps still aren't editable inside NotebookLM. Once a map is generated, you can't drag a node, rename a branch, or restructure the hierarchy without going back to the source material and regenerating from scratch. If the map comes back 90% useful, but one branch is mislabeled or sitting in the wrong place, your only option inside NotebookLM is to try again with a tweaked prompt and hope the next generation gets it right.

For now, my workaround is exporting the Mind Map and bringing it into XMind, which is better suited for deeper changes like adding branches or restructuring topics. It's an extra step, and it breaks the otherwise self-contained NotebookLM workflow, but it's the only way to actually edit a map without restarting the generation process.