NotebookLM doesn’t really need an introduction; it’s one of the best and most popular research assistants and learning tools out there. However, when you use it right out of the box, it can feel a tad generic. Even if you give it a very specific prompt, it might not give you the exact answer you’re looking for. Sometimes, it also assumes you know more about the subject than you actually do, so some of its responses don’t always click. I especially run into this issue when I’m researching topics I’m unfamiliar with.

And then NotebookLM introduced Custom Mode. This is almost like a system prompt for individual notebooks, allowing you to configure how the AI behaves and responds. It completely changed how I use NotebookLM and how much value I get out of it. Here’s how I’ve been using it…

What is NotebookLM's Custom Mode?

The feature that lets you mod NotebookLM

NotebookLM has been rolling out additional features throughout 2025; one of those is the Custom Mode. It’s a relatively new NotebookLM configuration setting that works a lot like a system prompt, but only applies to a single notebook instead of your entire NotebookLM account.

Just like a system prompt, you can give NotebookLM context about your skill level on a topic so it doesn’t overcomplicate explanations or assume prior knowledge. You can also instruct it on how you want information delivered - such as short bullet points, table formats, step-by-step explanations, or practical examples. You can even specify tone and grammar preference or things you explicitly don’t want.

I actually created an “AI version” of myself with NotebookLM and relied on the Custom Mode to do it. And in doing this, I’ve managed to get the AI to forgo all capital letters in its responses despite it not being proper grammar, which demonstrates how accurately this mode operates.

Custom Mode is especially useful when you use NotebookLM for very different purposes. One notebook might be for technical learning, another for writing, and another as a dumping ground for your journal entries. So instead of explaining the format and tone you want in every prompt, Custom Mode keeps those rules in place until you modify or remove them.

You can find the feature by going to the slider icon in the top-right of the chat panel. This will open a Configure Chat window. Under “Define your conversational goal, style, or role”, select Custom and enter your prompt.

How I make the most of NotebookLM’s Custom Mode

Getting personalized and consistent outputs

I use Custom Mode to save time and reduce any friction I used to have in my notebooks. So after adding my sources, but before I start prompting NotebookLM, I set a few rules based on what that notebook is for. This way, I don’t have to keep correcting responses or restating context.

For learning-heavy notebooks, I tell it my current skill level and how I want it to formulate its responses, followed by practical examples when necessary. I also ask it to avoid assumptions and skip filler. This is especially useful for coursework or more technical topics.

For example, here’s my current Custom Mode prompt for my UX design learning notebook:

You are helping me learn and refine UX design skills. Assume I already understand:

-Core UX principles

-User flows, wireframes, usability basics

Do not explain fundamentals unless I explicitly ask. Treat me as intermediate:

-I want sharper reasoning, not definitions.

-Skip “what is UX” explanations.

-If a concept is advanced, explain it through application, not theory.

Prioritize:

-Practical decision-making

-Trade-offs

-Constraints

Avoid:

-Generic UX advice

-“It depends” without follow-up

And this is my system prompt for my local LLM notebook where I’m learning how to self-host LM Studio:

-You are helping me learn how to run and use local LLMs via LM Studio.

-Assume I am a beginner to self-hosting, but not a beginner to general tech tools.

-Do not assume Docker, command-line tools, or server deployment unless explicitly asked.

For more research-oriented notebooks, I’ll use Custom Mode to modify structure. I’ll ask for bullet-point summaries, clear source attribution, tables of information, and short comparisons instead of long-form explanations. This makes it much easier to scan outputs and pull the information into an app like Obsidian later.

And for writing and more personal use cases, I configure NotebookLM to focus on clarity and consistency - short paragraphs, neutral tone, and no marketing language. That way, feedback is more grounded and usable.

When Custom Mode isn’t necessary in NotebookLM

Knowing when not to use it

Custom Mode isn’t needed for every notebook. If I’m just doing quick lookups, reading documentation, or pulling a fast summary, the default NotebookLM behavior is already fine. Setting up extra instructions in these cases just slows me down.

I also skip it for short-term notebooks or experiments where I don’t need consistent tone or structure. It’s best reserved for notebooks I plan to reuse or build upon.

Making NotebookLM work for me

NotebookLM’s Custom Mode is a powerful way to tailor its outputs and maintain consistency across a notebook. For long-term projects or specialized learning, it can be a game-changer. But it’s not mandatory for every single notebook. Using it selectively this way keeps my work efficient without unnecessary overhead.